Constructing the perception of ‘annoying’ words and phrases in interaction: An analysis of delegitimisation strategies used in interviews and online discussions in Finnish

3.1 Interviews and online discussions

In this article, how people respond to the question on ‘annoying’ words and phrases is examined in language biography interviews recorded during the research project One Hundred Finnish Linguistic Life Stories, conducted at the University of Helsinki from 2017 to 2020 (Priiki et al. 2018, Hippi et al. 2020). The predetermined questions in the interviews include the question Onko muiden kielenkäytössä piirteitä tai sanoja, jotka ärsyttävät sinua? Jos on, millaisia? ‘Are there aspects or words in the language used by other people that annoy you? If yes, what are they?’

The responses given to this question in 92 interviews were compared to two discussion threads (90 messages each) entitled Ärsyttävät sanat ja sanonnat ‘Annoying words and phrases’ and Mikä ärsyttää ihmisten kielenkäytössä? ‘What annoys you in people’s language use?’ on an anonymous online forum called Suomi24.fi.

These two threads were selected for the analysis, since the questions asked resemble the interview question as closely as possible. Spontaneously occurring discussions that include exactly the same question were not found. Of course, the formulation of the question affects the responses (van der Zouwen 2002, 50), and the main difference between the two data sets is that the online forum questions assume that respondents are annoyed by something in the language. The analysis revealed that, despite the formulation of the interview question, some interviewers appear to anticipate an affirmative answer.

In language biography interviews, narratives are constructed dialogically focusing on language acquisition, language use, and language attitudes. Language biographies may be collected as texts written by the studied persons themselves or through surveys and interviews. The present data are a cross-sectional sample of the population living in Finland and speaking Finnish either as their mother tongue or as a second (or third) language. The interviewees come from different social groups and regions, and they represent different linguistic varieties of Finnish and Finland–Swedish,^[2] as well as other minority languages in Finland. In the collected thematic interviews, the questions are related to the interviewees’ residential history, family, events in life, and perception of their own languages and linguistic skills, as well as their stance on different languages and varieties. The interview questions were designed in collaboration with a multidisciplinary research team, and the interview framework was based on previous sociolinguistic studies on Finnish (Hippi et al. 2020, 24). The framework aimed at creating a comprehensive picture of life stages, language use situations, and language perceptions of each interviewee, and the questions were designed to encourage the production of narratives (ibid.). The author of this article was part of the research team.

Interviews are a central data collection method in various fields, which is also true for sociolinguistics. Earlier, interviews could be uncritically regarded as a source of accurate information; however, in recent decades, there has been a growing understanding that an interview is an interactive situation influenced by a variety of factors (Silverman 1997, 248, Atkinson and Silverman 1997, Wengraf 2001). An interview stripped from power relations is an impossibility: the interviewer controls the interview and the researchers its interpretation; however, the interviewee has the power to choose what to say and how to say it (Holstein and Gubrium 1995, 4, Kvale 2007, 483–6). Therefore, neither party is a passive transmitter of information; however, interviews are active, changing situations, even if the format of the questions is fixed and the interviewer attempts to influence its course as little as possible (van der Zouwen 2002, 51–2). Earlier studies have shown that, for example, face-to-face interviews yield more complete responses than mail surveys and that an interviewer’s social background has only a limited effect (ibid. 51). However, interviewers’ behaviour is an important factor affecting responses given (ibid. 52).

The interviews analysed in this study are semi-structured biographic-narrative interviews (for the aspects of this type of interview in general, see Wengraf 2001, 5) and they were video- or audio-recorded. Most of the interviewers were university students participating in a project course. They were trained in the task, but for most of them, this was their first experience as an interviewer and they only conducted one interview. Thus, most interviewers in the data are inexperienced. The interviewers had slightly different practices in, for example, how they formulated the predetermined questions, how they linked the question regarding annoying language to a previous discussion during the interview, and whether they asked follow-up questions. These are typical aspects of an interviewer’s behaviour that can affect the comparability of the data (van der Zouwen 2002, 60–1). As Wengraf (2001, 5) indicates, in semi-structured interviews where interviewees are asked to produce narratives, interviewers need to improvise their responses; this makes this a challenging interview genre.

As mentioned earlier, interviews are interactive situations where the actions of both the interviewer and the interviewee influence the manner in which the conversation unfolds. In the current data, the manner in which participants answer questions is affected by the order of the questions presented, as the topics currently discussed can lead to the discussion of other topics. Occasionally, the interviewers explicitly related the question in focus here to topics already discussed. The relationship between the interviewer and the interviewee also plays a role. Most often, the participants were not previously known to each other, but the people being interviewed may make a few assumptions regarding the interviewer’s language ideologies based on their knowledge that the interviewer is, for example, a young female university student in linguistics; this, of course, may affect their responses, even though the effect of the interviewer’s social background appears to generally be a limited one (van der Zouwen 2002, 51). Some assumptions are also embedded in the interview questions, and some may be revealed by the interviewer’s reactions to the interviewee’s answers.

Although conversation analysis has its origins in the study of everyday interaction, nowadays, it is widely used to study institutional situations (Drew and Heritage 1992). Videotaping the data and analysing interview interactions using the methods of interactional linguistics and conversation analysis help to identify the problems and benefits of the formulation of interview questions and the interviewer’s behaviour. As mentioned earlier, the course of a research interview is influenced by a wide range of interactional factors. To a certain extent, these follow the same logic as everyday conversation (Wengraf 2001, 3), and practices that work well in everyday conversation appear to also be most effective in interview situations (van der Zouwen 2002).

Written online conversations, to which the language biography interview data set is compared in this article, are very different from interviews in a communication context. Online discussion forums are a form of asynchronous interaction, thereby differing from chat rooms and instant messaging applications. These are multi-party interactions where responses are often delayed and participants use temporary pseudonyms instead of their real names. Online discussions are open to a wide range of participants, and they are considered as unregulated linguistic spaces that allow heterogenic language use. Despite greater visibility of written nonstandard varieties, social media does not appear to contribute to the acceptance of linguistic variation. On the contrary, in social media contexts, participants – driven by standard language ideology (see Milroy 2001) – often reinforce norms by practising verbal hygiene (Sherman and Švelch 2015, Heuman 2020).

There has been a growing interest in discourse practices in online interaction from the beginning of the twenty-first century. Online discourse forums have been analysed by studying opening posts (Antaki et al. 2005), responses to them (Vayreda and Antaki 2009), and turn-taking practices (Gibson 2009). Furthermore, Finnish asynchronous online discussions have been examined using conversation analytic methods by Virtanen and Kääntä (2018) and Määttä et al. (2020). Asynchronous written online discussions often consist of rather long and complex posts and, as they are usually read by a larger audience than only the active contributors of the discussion, the texts may address an assumed public audience (Virtanen and Kääntä 2018, 152). Sequences in discussions can be organised non-chronologically, but participants still treat them as adjacency pairs, such as a question and an answer (Gibson 2009). Because written discussions lack the cues, such as glances, that face-to-face conversations use to regulate turn-taking, applications often utilise technical means to, for example, mark a message as a reply to someone else’s post. Online conversations are today inherently multimodal, using emojis, images, and videos as part of the communication (Thurlow et al. 2020). Unfortunately, images posted in threads are not available in the online data used in this study. The forum and topic in question, however, are old-fashioned in the sense that the debate is text-based.

The two discussion threads examined in this article were taken from the discussion forum Suomi24.fi, which is open for research purposes via the Language Bank of Finland. The forum data are introduced in detail in Lagus et al. (2016). The open and anonymous forum comprises writers and readers, who may be casual browsers or active contributors, and is moderated on the basis of requests for deletion (Lagus et al. 2016, 9). The Suomi24.fi forum has a diverse user base compared to, for example, closed Facebook groups (Harju 2018, 52), where language issues are also discussed.

I refer to the discussion titled ‘Annoying words and phrases’ as Discussion 1 and to the discussion titled ‘What annoys you in people’s language use?’ as Discussion 2. Added to the declarative title, the opening post of Discussion 1 includes the question, ‘Which words make you furious or at least annoy you a lot?’; as examples, the writer provides a list of six expressions. The opening message of Discussion 2 frames the question to spoken language by asking people to ‘List things that annoy you in how people talk’. This opener also provides examples of linguistic aspects that annoy people. These discussions are interactive by nature, as participants form adjacency pairs by answering the opening questions or responding to the directive action. In addition to answering the question and producing the requested lists, participants discuss the aspects mentioned by others, agreeing or disagreeing with their perceptions.

3.2 Processing the data

The question about annoying features was asked in 92 of the examined 103 interviews. Annoying aspects were discussed in connection with other questions as well; however, to strengthen the comparability of the data, the analysis here is limited to the answers to this question only. When a particular phenomenon being annoying is discussed in the context of another question, this topic inevitably influences the interpretation. The examined parts – that is, the interview question regarding annoying language in each of the 92 interviews, the interviewee’s answer to it, and any follow-up questions by the interviewer – form a data set of ca 5,220 words of transcribed text.

The two discussion threads, to which the interview data set was compared, were found by searching with words ärsyttävä – ‘annoying’, sana – ‘word’, and kieli – ‘language’. To create a sample of approximately the same size as the interview data set, the first 90 posts of each discussion thread were saved. The size of the sample from Discussion 1 is ca 2,290 words, and the sample from Discussion 2 is ca 4,370 words. Discussion 1 was opened in January 2015, and the last of the examined messages was sent in February 2016, which is over a year later. Discussion 2 was livelier: it was opened in June 2013 and the first 90 messages were sent in just 11 days. Most of the replies are short, with the average length being 37 words.

The data are processed using content analysis, which identifies patterns in a systematic manner by focusing on the language and content, as well as the contextual meaning of a piece of text (Riffe et al. 2019, 20–35). Content analysis refers to a group of analytic approaches with different emphases, ranging from the simple counting of words to intensive processes of interpretation (Hsieh and Shannon 2005). The analysis conducted for this study begins with the form of content analysis which is characterised as conventional (see id. 1279–81). The data are examined to determine which linguistic features are explicitly mentioned as annoying in the interviews and online discussions and to determine the relative frequency of these mentions. In the analysis, each response to the interview question, a turn of speech, and each message posted to the examined online discussions function as a content unit. Keywords are identified from these units and categorised under recurring themes. The keywords are different levels of language-related expressions, such as metapragmatic references to words, phrases, varieties of language, or ways of speaking. For example, the mentions of words niinku (‘like’), tota (‘well’), and elikkä (‘so’) are categorised under the category ‘filler words’.^[3] The analysis is not, however, purely conceptual in the sense that I had to identify when the mention was a metapragmatic observation and not, for example, a pragmatic marker as part of the interaction. Some content units (online forum messages or response turns) are labelled under several categories, since longer messages and responses include mentions that fall into more than one theme.

The analysis is then continued using directed content analysis (see Hsieh and Shannon 2005, 1281–3), where units are labelled according to van Leeuwen’s (2008, 105–23) (de)legitimisation strategies. Because describing something as annoying is a moral judgement by itself, units in which there is no other justification for the perception are labelled in the category of ‘moral evaluation’. In van Leeuwen’s model, the strategy of authorisation is divided into personal, expert, role model, and impersonal authority, authority of tradition, and conformity. Only the latter three are found in the examined data. In this coding, impersonal authority refers to rules concerning language use that are learned at school or home, and authority of tradition refers to established practices regarding how one should use language in a specific context. I distinguish them from the authority of conformity, which refers to the use of the widespread nature of the phenomenon as a justification, since in these data, these strategies are used in different ways (Section 5). The strategy of rationalisation is divided into two subcategories, as done by van Leeuwen (2008, 113–7): instrumental and theoretical rationalisation. Instrumental legitimisation invokes the practicality of a phenomenon, while theoretical legitimisation appeals to its logicality. Van Leeuwen’s mythopoesis refers to myths and storytelling as argumentative strategies. Here, cases in which the interviewees or online discussion participants justify their perception by reporting some interconnected events linked to a specific time, place, or experience are coded as narratives.

5.1 Moral evaluation

Delegitimising by moral evaluation includes claims that a particular use of language is bad, stupid – or annoying. People delegitimising language use to place themselves morally above others by stating that although a particular practice may be widespread or accepted by an authority, they would not use it themselves based on their values. Note that, in this dataset, the questions under consideration are about annoyance, not what kind of language is perceived as ‘good’ or ‘bad’.

In example 1, a 28-year-old woman justifies her feeling of annoyance with physical sensations: in addition to irritation, her ears ache and she gets an unpleasant feeling when she hears swearing. In the interviews, swearing is the most frequently mentioned annoying aspect and is typically delegitimised by moral evaluation.

(1)^[4] LN1989

In this dataset, different aspects of language cause different levels of reactions. Example 2 from Discussion 1 illustrates well what most of this thread is like: messages list examples of annoying words and phrases without explaining them in detail.

(2)^[5] Discussion 1

The words listed in message 15 of example 2 are common everyday colloquialisms, which are not usually mentioned in the lists of ‘annoying’ words. The degree of annoyance for them is, in the writer’s opinion, only ‘somewhat annoying’. Only in the case of the word homo ‘gay’ do they give a reason other than an emotional reaction. I address the appeal to logic in the next subsection. Message 16, in contrast, shows the difficulty of interpreting online discussions: it is probably intended to be an example of annoying language (the message refers to itself as ‘shitty Finnish’), but the negative assessment may be directed at faltering spelling, the use of capital letters or word choice. Message 16, on the other hand, relies on the description of an emotional reaction stronger than annoyance. The Finnish generic you, which is a rather typical source of irritation (Joronen 2007, 45), drives the writer ‘to the brink of rage’.

The examples in this section illustrate how describing language as annoying is a well-established way of talking about it, and in both interviews and online discussions, appealing to an emotional reaction is usually accepted as a sufficient strategy for delegitimising both phenomena that are often perceived as annoying and those that are rarely disliked. Other participants in online discussions or interviewers rarely ask follow-up questions regarding why a phenomenon is annoying, considering the feeling to be a sufficient justification. The examples presented above also show that different aspects of language that discussions seek to delegitimise evoke different strengths and qualities of emotion.

5.2 Rationalisation

Delegitimisation by rationalisation involves invoking either the instrumental value of language or the common knowledge of what is reasonable and logical. An example of rationalisation in the data has already been shown above in example 2, where the writer of message 15 delegitimises the use of the word homo ‘gay’ as a term of abuse because they consider it illogical. This falls into theoretical rationalisation, as it claims that words should have a logical, unvaried meaning.

Instrumental rationalisation refers to the goals and effects that certain linguistic practices have as reasons for delegitimising them. Loanwords or strong dialects annoy the discourse participants because they are sometimes difficult to understand, and filler words distract the listener from the main point. In the data, the strategy of instrumental rationalisation is typically associated with demanding clarity of the message. Successful communication is, then, seen as the most essential function of language.

In example 3, the interviewee, who is a 24-year-old woman, first answers the question about annoying features that loanwords taken directly from English, and a spoken variety, the slang of Helsinki region ‘grate on her ear’. This variety of language is commonly given a negative stigma (Paunonen 2000, 42–4). As a further justification for the annoying nature of slang, N1993 mentions that it is difficult to understand. The example shows how different strategies are combined, as already done in example 2.

(3) N1993

In the current data, online Discussion 2, in particular, deals with grammatical errors. A typical strategy for delegitimising incorrect grammar is theoretical rationalisation, which appeals to logical reasoning or to ‘the ways things are’. Grammar is seen as a system that should be logical and symmetrical, even though natural languages always contain certain irregularity. This reflects the standard ideology aiming at reducing language variation (Milroy 2001, Walsh 2021). The admiration of a logical system was evident in the development of the Finnish written language in the nineteenth century: some language developers at the time proposed forms that did not appear in the vernacular to make the paradigms systematic (Nordlund 2004, 300).

Before example 4, the participants of Discussion 2 debated the correct inflection of the proper name Thaimaa (‘Thailand’). The non-standard illative form Thaimaaseen (resulting from the analogy to other names, such as the lake Saimaa) is relatively widely used instead of the correct form Thaimaahan (Maamies 1996), which annoys some people (Joronen 2007, 42). Message 32 introduces another name into the debate. Sörnäinen refers to a district of Helsinki; its illative form in standard Finnish is Sörnäisiin, which, as the writer of message 32 indicates, appears as a plural form. Many Finnish place names ending in -nen have such inflections (Salonen 2009). The writer legitimises the more logical singular form Sörnäiseen by appealing to the fact that there is only one district with this name.

(4) Discussion 2

In this subsection, I have shown that the strategy of rationalisation in the data delegitimises both whole language varieties (e.g. by appealing to their obscurity) and the meanings and inflections of individual words. Rationalisation is generally used to support moral delegitimisation, but illogicality or instrumental weakness can also be a reason in itself to delegitimise a feature of the language, as if for the common good.

5.3 Authority

As mentioned, the data contain instances where the delegitimisation of a linguistic feature is justified by referring to impersonal authority, tradition, or conformity. A typical impersonal authority found in both the interviews and online discussions is school education. Language users base their understanding of what constitutes legitimate language use on the norms of language they have learned in school.

As Sherman and Švelch (2015) indicate, the forms perceived as correct do not always correspond to the current recommendations provided by an official institution. This is evident in the present data: in example 5, M1954a, a man in his sixties, refers to the rule about the Finnish construction tulee tehdyksi ∼ tulee tehtyä ‘get done’. In this construction, the Finnish Language Office preferred the variant in which the translative case is used (tehdyksi) until 1973, but then the variant with the partitive case (tulee tehtyä) was accepted in standard Finnish (Kielitoimiston ohjepankki: Passiivimaisia rakenteita: tuli tehtyä ∼ tehdyksi).

(5) M1954a

In this example, M1954a first gives a negative answer to the interviewer’s question (turn 4). However, the interviewer brings up an issue mentioned by M1954a earlier in the conversation (the hyphenation), in an attempt to activate an answer in the interviewee’s mind. M1954a still denies this, explaining that even he is unable to follow the ideal of good language that he has in mind (‘my owns rules that I have learned at school’). As an example of this ideal language, he gives the structure mentioned above. M1954a still considers the partitive form incorrect in this structure because it was viewed as such in his school days, even though the standard has changed. He also refers to the spoken language, although the Finnish standard language is generally only used in writing and denies the authority of conformity (‘everybody does it’) by stating that being widely used does not make the form acceptable. Despite first responding that he is not really annoyed, in turn 12, he admits that the ‘incorrect’ version annoys him ‘like hell’.

Authority is not only referred to when discussing standard language, but questions about legitimate forms of language are also raised considering dialects. In the interview data, some people are annoyed by others that use a regional variety ‘incorrectly’. In example 6, the interviewer asks a question that is followed by a pause of several seconds. The interviewee, M1957, a man about 60 years of age, does not answer right away. The delay prompts the interviewer to provide examples of potentially annoying aspects: she mentions slang and colloquialisms. Even after this, the interviewee thinks for several seconds, begins with a negation (no ei see ‘well, not that’), delays, and hesitates. Ultimately, he produces a response (turn 2) stating that he gets annoyed if slang or dialect is used by someone who ‘doesn’t master it’. Hesitation indicates that he finds it difficult to formulate his ideas or that he is unsure about the relevance of the answer.

(6) M1957

After the interviewee’s explanation, the interviewer produces a kind of candidate understanding (see Antaki 2012) of what she thinks M1957 means by the misuse of dialect. M1957 mentions the non-standard illative form Turkkusee (‘to Turku’), which is a widespread dialect imitation but not used in actual dialect (Mielikäinen 2005, 105), earlier in the interview when they discussed Finnish dialects. The authority for determining the correctness of a regional variety is probably a representative of the oldest, ‘original’ and ‘purest’ dialects possible. This reflects the traditions of Finnish dialect research (Aarikka 2023, 155–6, 440–2).

The authority of tradition and impersonal authority are not always easily distinguishable in the data, where language change is often seen as something to be resisted. Although both dialects and official standards change, language users tend to see older forms as better. Authority of tradition also invokes the idea of established practices of how language should be used in a specific context, for example, in a workplace. Language users, both in interviews and on the discussion board, agree that negatively perceived linguistic phenomena, such as swearing, are more problematic at work than during leisure time. In example 7, the writer of message 13 is annoyed by the particle elikkä ‘so’ used by a shop assistant. Message 14 verbalises the rule of situational variation: ‘you should talk properly at the checkout, no matter what you do between friends’.

(7) Discussion 2

Example 7 shows also how, in online discussions, some participants intend to restrict the participation of others by delegitimising language use. Message 15 is written with several orthographic errors, which is ironic because the writer says that they are annoyed by the sloppiness in language and the incorrect use of capital letters, simultaneously revealing the same aspects in their own writing. In message 16, another participant corrects the orthography. If a writer makes typos in their own post, their opinion will be invalidated on the grounds that they cannot demonstrate their knowledge of standard language rules. This is not surprising, since good control of linguistic norms and orthography is connected to morals, intelligence, and higher education (Niedzielski and Preston 2010, 224). This kind of practices may limit the participation of L2 speakers or people with low education levels in the discussion.

Anonymous public online forums, such as the discussions examined here, have a wide audience of potential readers, and the participants may aim to entertain this audience (Virtanen and Kääntä 2018, 152, see also Marwick and Boyd 2011). Offensive language is the downside of online anonymity and freedom of expression (Kim and Herring 2018), and the so-called ‘trolling’– which is deceptive and aggressive behaviour intending to provoke other online users – has become an inescapable part of online communication (Ortiz 2020). Therefore, it is occasionally difficult to tell how seriously the participants in these discussions are taking the debate and how reliable the information they provide is. Some messages might be deliberately written carelessly to create an ironic impression, with an intention to provoke or entertain others. Considering message 15 of example 7, shown above, the non-ironic nature of the message is supported by the impression that the same participant returns later in the thread by blaming their malfunctioning keyboard for typos.

5.4 Narratives

Van Leeuwen’s fourth (de)legitimising strategy, mythopoesis, refers to storytelling; this seems to be quite infrequent in discussions on annoying linguistic phenomena. Even though the interviews were designed to encourage the interviewees to tell stories about their linguistic experiences (Hippi et al. 2020, 24), the question about annoying features does not often prompt narratives as responses. Instead, the interviewees describe their perceptions on a general level. However, a few instances in which a participant shares an incident of unsuccessful language used as a cautionary tale are found both in the interviews and in the online discussions.

In example 8, the interviewee, a woman about 30 years of age, answers that she is annoyed by swearing and filler words in general. Then, she gives an illustrative description of a situation in which she is sitting on a train, hears other people talking, and starts counting the number of fillers. N1988 brings the story to life with a direct quote on her own thoughts (turn 8).

(8) N1988

Storytelling is not common in online discussions either, but a few examples were found. In example 7, the writer of message 13 links their feelings of annoyance to the time and place of ‘in Pori in Lidl yesterday’. A longer and more detailed narrative is found later in the same thread (Discussion 2). There, a participant is annoyed by the use of the distal demonstrative pronoun to refer to themselves. Using this pronoun to refer to someone present in the situation is generally considered somewhat impolite by Finnish speakers (Priiki 2015, 44–6). The writer uses the story as an example of rude language.

(9) Discussion 2

In general, it is rude to talk about strangers so loudly that they can hear. In the discussion that follows message 56 in Discussion 2, it is clear that the writer draws attention to the word toi ‘that’ in particular. For example, after the story, they invoke common knowledge and tradition by writing, ‘But of course, all educated people know anyway that people are not spoken of in the style toi’.

Narratives are, therefore, a rare delegitimisation strategy in this dataset, but they can be used as cautionary examples to illustrate what kind of language use is meant and what consequences it may cause.

5.5 Counter-reactions to verbal hygiene

All examples presented above show strategies of delegitimising rather than legitimising language use; that is, people use them to justify disapprovement. As I have already mentioned, especially in online discussion 2, there is a genuine debate, though, where some participants also defend the linguistic features disapproved of by others. It can be seen in this discussion that rather than seeking consensus by explaining their arguments in an understandable manner, many messages are intended to pick on and verbally beat the other participants (about online discussion culture and ‘trolling’, e.g. Kim and Herring 2018, Ortiz 2020). This kind of behaviour typical of anonymous online interactions is questioned on a meta-level. A comment in the thread explicitly directs the participants to focus on answering the question rather than mocking others: “Instead of giving your assessment, why don’t you tell us what you think about the original issue, but comment negatively on the opinion of others?” (Discussion 2, message 69).

Discussion 2 contains many counter-reactions to annoyance expressed by others. In Discussion 1, though, only one message explicitly condemns the debate; this is presented as example 10. Message 12 in example 10 addresses other participants in the debate, questioning the annoying nature of the words listed in previous messages and stating that most of them are part of ‘normal’ language use, making use of legitimisation by the authority of conformity.

(10) Discussion 1

Message 12 does not elicit explicit reactions and the discussion continues with listings, similar to message 11. However, message 13, following the questioning post, is ambiguous, as it is technically marked as a reply to the previous message. It is possible to interpret the two phrases it contains as an address to the writer of message 12 (meaning ‘Really, retard?’). On the other hand, in message 13, the writer may simply continue listing annoying words by stating that they are annoyed by the expressions aikuisten oikeesti and vajakki.

The resistance found in the examined Finnish online discussions is contrary to what Heuman (2020, 10) reports from Swedish discussions on the Jodel app. In example 11, counter-arguments are presented by asking rhetorical questions and quoting and echoing others’ messages. The discussion continues with the same theme presented in example 4 (in Section 5.2), the incorrect inflection of the word Thaimaa ‘Thailand’ in the illative (about this, see Maamies 1996). This aspect is widely debated in Discussion 2, where the nonstandard inflection form is also defended.

(11) Discussion 2

The difficulty for Finns in inflecting the word Thaimaa (‘Thailand’) is that because it is a name, many people do not perceive it as a compound word with the word maa ‘land’ at the end. When understood as a simple word, it is inflected by adding the same type of illative ending as other place names ending in a long vowel, such as Lake Saimaa and the city of Espoo, which are mentioned as analogies in example 11. The writer who raised the issue in Discussion 2 mentions the inflection of the word maa (‘land’) in their post; this is quoted in the beginning of message 37. Those who defend the nonstandard inflection refer to the analogy of other names (e.g. Saimaaseen and Espooseen). The rhetorical question presented in message 37 is a response to message 29, which is evident from the quote given in the message. The question is not answered, but prompts a counterargument in the form of another rhetoric question (message 38). The next message again quotes the beginning of the message that initiates the discussion regarding the form Thaimaaseen and echoes the same structure, providing analogies to defend the non-standard inflection.

In interviews, the interviewer’s role is not to assume a position on whether they agree or disagree with the interviewee, and in these data, the interviewers stick to this fact. However, in a few interviews, the interviewees themselves offer counter-arguments to their experience of irritation and reflect on different aspects of the phenomena. In example 12, N1957b – a woman about 60 years of age – begins a long explanation after her negative answer.

(12) N1957b

First, she explains that the perceived ‘ugliness’ is related to the content of talk and not to the linguistic forms themselves. In contrast, she understands the reasons why young people swear, even if she does not like it. Similarly to M1954a (in example 5 in Section 5.2), she admits that the annoying phenomenon belongs to her own language, appealing to the authority of conformity. She also understands swearing as having a function among young people, which legitimises it by instrumental rationalisation. Her interpretation of the reasons for young people swearing is marked with laughter, which marks the subject as a somewhat extraordinary and contradictory one (see Haakana 2001).

Both the online discussions and the interview data show counter-reactions to verbal hygiene practices. Online, it is a genuine debate with disagreements, while in interviews, people reflect on different aspects of the phenomenon that annoys them and show an understanding of its causes and functions. Different strategies are used to legitimise phenomena from those used to delegitimise them: in their counter-arguments, people appeal to the authority of conformity or to instrumental rationalisation.