A cognitive-semiotic approach to impoliteness: Effects of conventionality and semiotic system on judgements of impoliteness by Russian and Swedish speakers

2.1 A cognitive semiotic framework

One of the goals of cognitive semiotics is to provide insights into complex phenomena like meaning, consciousness, intersubjectivity and conventionality, and more particularly, on how these are realized in semiotic systems such as language, gesture, music, and depiction. The first two of these systems are key to our study and are discussed below. Methodologically, cognitive semiotics aims for well-balanced investigations, combining conceptual and empirical aspects (Konderak 2018; Zlatev 2015a). For this purpose, it utilizes phenomenological triangulation, which, unlike standard methodological triangulation, is predicated on the integration of methods from three different perspectives. The first-person perspective focuses on the direct experience of a given phenomenon, using methods like phenomenological reduction and intuition-based analysis (Gallagher and Zahavi 2012). The second-person perspective is that of the other person, and it zeros in on social interaction, emphasizing that participants in all studies are human beings with their own experiences. Only after applying these two perspectives may the third-person perspective of detached observation and measurements be applied. Recent studies applying this approach have focused on diverse phenomena such as polysemiotic narratives, choice awareness, and street-art metaphors (Louhema et al. 2019; Mouratidou 2019; Stampoulidis 2021).

There are many different theories concerning the nature of signs (and other meanings like signals, affordances, categories, etc.), but the one we adopt is based on the model known as the Semiotic Hierarchy (Zlatev 2009, 2018). In short, this states that meaning-making (semiosis) and intentionality, in the phenomenological sense of the directedness of consciousness towards the world, are two aspects of the same subject-world relationship. In sensory-motor activity, perception, and imagination there is no sign use but only an intentional link between the mind and the respective intentional objects. Sign use becomes possible only with considerable reflective consciousness, allowing the subject to know that a given expression (in some particular medium, see below) denotes a given intentional object, which could be a thing, a person, a property, a whole event, or even an attitude (like impoliteness). It is this reflective awareness of the signitive relation between expressions and intentional objects that distinguishes signs like words and gestures from signals like spontaneous face expressions (Zlatev et al. 2020).

Neither signs nor signals operate on their own, but form systems – combinations of signs/signals of a particular kind, and their relations. Language – realized as speech, signed language or writing – is a universal and paradigmatic human sign system. However, it is by far not the only one, given that all known cultures make extensive use of expressive movements of the hands and the rest of the body for the purpose of signification: gesture (Kendon 2004). These two sign systems differ fundamentally in terms of their materiality (i.e., how they are produced and perceived) and semiotic-structural properties, as shown in Table 1. A third semiotic system, not investigated in the present work, is that of depiction: the production of marks on 2-dimensional surfaces, serving as iconic representations of 3-dimensional objects (Zlatev 2019).

In terms of medium, language is similar to gesture only in the case of signed languages, given that both are produced by the body (see Table 1). In terms of (perceptual) modality, gesture and writing are perceived primarily visually. But the original and still most frequent form in which human language is realized is that of speech, which utilizes the auditory modality. Turning to the semiotic-structural properties, language has double articulation: phonemes or graphemes combine systematically to form meaningful morphemes, and this is also the case for some but not all signed languages, in particular those that have emerged more recently (Sandler 2012). In terms of semiotic ground – the type of relation between expression and object – in language this is predominantly conventional, even if iconicity (resemblance) and indexicality (spatio-temporal contiguity) are also relevant (Sonesson 2007). The syntagmatic (sequential) relations between its signs are characterized by a high degree of compositionality, where the meaning of a composite sign is built up (at least in part) from the meanings of its constituent signs, and the rules for combining these. The signs of languages fall into two broad categories, syncategorematic and categorematic, together allowing language to express messages of high internal complexity, making it considerably more precise than other sign systems, including gesture (Bundgaard 2010; Sokolowski 2008).

Gestures are bodily actions performed deliberately and perceived as expressing some “meanings” (or, in our terms, intentional objects) rather than for the sake of practical aims (Kendon 2004; Zlatev 2015b). The signs of gesture can be analysed into units, phrases, and nuclei, but these are not made up of minimal distinctive elements like phonemes or graphemes, and hence lack double articulation (Green 2014; Kendon 2004). Further, the sign system of gesture has much fewer systematic manners in arranging sequences of signs, making it more difficult, though not impossible, to express complex messages such as narratives. The predominant semiotic grounds of gesture are iconicity and indexicality, even if conventionality is also important, especially in so-called ‘emblems’ like the ok-gesture or the raised middle finger. Gestures can be combined in a linear sequence to express a compound message, and even simpler narratives, as in pantomime (Zlatev et al. 2020). These, however, lack the articulate structuring of language and are inherently limited in semiotic complexity.

How did sign use emerge in human cultures, given that no other species on the planet possess this capacity (Deacon 1997; Hurford 2007)? A plausible explanation is that the first human semiotic system based on signs rather than signals was that of pantomime: with highly iconic enactment of whole-body gestures at its core, but complemented by vocalizations and facial expressions (Zlatev et al. 2020). This proposal is based on the influential theory of human cognitive evolution developed by Donald (1991, 2001, according to which over two million years ago our ancestors evolved the capacity for bodily mimesis, realized in the ability to control the movement of the body to an unprecedented degree, allowing high-quality imitation, tool production, pedagogy, and intentional body-based communication, i.e., gesture. Over the following two million years, bodily mimesis was gradually complemented by vocomimesis, which with processes of cultural evolution and conventionalization gave rise to speech and grammatical language (Collins 2013; Heine and Kuteva 2007). However, gestures are still universally used in human communication, making it together with language (and other semiotic systems) fundamentally polysemiotic (Louhema et al. 2019; Zlatev 2019). While gestures differ extensively in terms of degrees of conventionality and iconicity, their basis in bodily mimesis and “a general mechanism of action generation” (Kita and Özyürek 2003: 30) makes them more “action-like” than language. Thus, it is possible that especially negative pantomimic gestures, i.e., those that are close to actual physical actions, are experienced as more threatening, and at least in some respects as more impolite than verbal signs. To some extent, this may apply to all gestures given their origin in physical actions (Müller 2016). This applies even more to our study, where we took a broad notion of “gesture”, including communicative actions like opening a door for someone (as a sign to leave), or pushing someone’s belongings off the table.

2.2 Approaches to impoliteness

The field of (im)politeness studies is diverse and highly interdisciplinary. Although this has contributed to some disagreements on conceptual issues, “a widening of the paradigm” is undoubtedly a positive outcome (Culpeper et al. 2017: 6). By implementing our cognitive-semiotic framework and combining it with existing approaches, we intend to provide new insights into impoliteness as a universal human phenomenon. Two distinctions that we adopt are those between positive and negative face (Brown and Levinson 1978, 1987) and between first-order and second-order politeness (Eelen 2001; Watts et al. 2005 [1992]), thus integrating a cross-cultural perspective, as well as scientific and everyday conceptions of (im)politeness. Further, these distinctions go well in line with the phenomenological triangulation of cognitive semiotics presented above.

Nearly all of the plethora of definitions of impoliteness include exhibiting emotionally negative behaviour towards someone or something. A definition accommodating such factors as emotions, deliberateness, and expectations, which fits the purposes of the study, is the following:

Impoliteness is a negative attitude towards specific behaviours occurring in specific contexts. […] Situated behaviours are viewed negatively when they conflict with how one expects them to be, how one wants them to be and/or how one thinks they ought to be. Such behaviours always have or are presumed to have emotional consequences for at least one participant, that is, they cause or are presumed to cause offence. Various factors can exacerbate how offensive an impolite behaviour is taken to be, including for example whether one understands a behaviour to be strongly intentional or not. (Culpeper 2011: 23)

Despite being ubiquitous, impoliteness is less frequent than polite or politic behaviour in everyday social interactions (Culpeper 2011; Watts 2005). Quite often, “the interactant who utters impoliteness must have felt sufficiently provoked at some point prior to actually [deliver] the impoliteness”, but even when impoliteness is justified, these “justification thresholds” may vary among speakers (Bousfield 2008: 183). Further, impoliteness can be triggered not only when the speaker’s intentions to offend are clear, but also when they are falsely attributed to the act by the hearer; in the latter case impoliteness functions as an unintentional act.

One of the central concepts to our study is that of face. Originally, it was defined by Goffman (1967: 5) as “the positive social value a person effectively claims for himself by the line others assume he has taken during a particular contact”. A self-image that a person has (i.e., face) may be consistent or inconsistent with a particular context. When consistent, the self-image aligns with how others see it (e.g., I see myself as a distinguished writer, and others do as well). On the contrary, when there is a conflict between the internal and external images, a person may feel either uplifted when others attribute more positive characteristics to them than expected (e.g., I don’t see myself as a good speaker, but others do), or one can feel bad if others ascribe unexpectedly low value to them (e.g., I see myself as a good speaker, but others don’t).

The concept of face was later revisited by Brown and Levinson (1978, 1987 who made a distinction between positive and negative face: “Positive face: the want of every member that his wants be desirable to at least some others. Negative face: the want of every ‘competent adult member’ that his actions be unimpeded by others” (Brown and Levinson 1987: 62, emphasis ours). Their definition of positive face aligns with the original definition of Goffman as it stresses the desire to be approved, liked, and ratified. Negative face, however, emphasizes autonomy of an individual and unwillingness to be imposed upon. This definition, while leaving out such features as social interdependence or substituting them with more individualistic “member’s wants” (Culpeper 2011: 25), highlighted the mutual interest of people to negotiate, maintain and sustain each other’s face, rather than losing or damaging it (Brown and Levinson 1987; Goffman 1967).

Together with the two types of face, Brown and Levinson proposed corresponding face threatening acts (FTA) – utterances that threaten someone’s face either verbally or non-verbally.^[1] In the context of politeness, positive face may be typically enhanced by establishing common ground, friendliness, and giving compliments. However, an attack on someone’s positive face can be done through FTAs, such as open criticism (e.g., “this is badly written”), derogatory expressions (e.g., “you are not that smart”), and insults (e.g., “you asshole”). Negative face, on the other hand, is often concerned with minimising the imposition, which is achieved by using modal verbs or not exercising power in a direct manner. Thus, one could challenge someone’s negative face by restricting their freedom of action, using dismissals (e.g., “get lost”), making people do something against their will (e.g., “I’ve told you to bring it now!”), and showing utter indifference (e.g., “I don’t care”).

Besides its utility for distinguishing different kinds of (im)politeness, the distinction between positive and negative face has been repeatedly used to pinpoint cultural differences. In our study, we adopt the idea of seeing the distinction between two types of face as a continuum rather than a dichotomy (Jay 2016; Upadhyay 2010). This may result in a cline where cultures gravitating towards negative face favour politeness strategies that maintain “one’s personal space”, whereas positive face cultures value strategies used for maintaining affection and solidarity (Barros García and Terkourafi 2014: 2).

Some evidence suggests that Russian culture is predominantly oriented towards positive face. Comparisons between Russian and English linguistic cultures show that the role of negative face in Russian is less significant, which could explain the frequent use of direct instructions and requests (Takhtarova 2015). At the same time, the role of solidarity is influential as expressions showing closeness and familiarity are quite frequent (Annin 2010). Ogiermann’s (2009) study on apologising concludes that Russian culture exhibits features of collectivism and hence demonstrates preferences for positive face. Swedish culture is harder to classify in terms of face and “seems to have two opposing tendencies: one towards individualism and the other towards collectivity” (Daun 1991: 165). A collectivistic component emerging from the idea of social equality and justice contrasts with the acknowledgement of privacy and independence. A comparison between American and Swedish cultures suggests that the latter has a stronger inclination towards negative face (Kiesling 2015), although Swedish culture in some cases is considered less individualistic than American or English culture (Dittrich et al. 2011). Thus, one can hypothesize that if a culture gravitates towards negative face, attacking negative face will be perceived as more impolite than attacking positive face, and vice-versa for cultures where positive face dominates.

The politeness theory of Brown and Levinson nourished a multitude of research focusing on face, universalities, and speech acts. With the appearance of the discursive approach to politeness, the focus was shifted to the analysis of the individual’s conception of (im)politeness and how it unfolds in the actual discourse (Culpeper and Hardaker 2017; Jucker and Staley 2017). This approach also implies that (im)politeness is not inherent in expressions, as even giving an order or calling a friend an offensive word does not automatically lead to construing the expression as impolite (Fraser 1990; Locher and Watts 2005). Thus, (im)politeness should be seen “as social practice and as social interaction” (Mills 2017: 45). In order to differentiate between everyday and academic conceptions of (im)politeness, a distinction between first-order and second-order politeness was drawn (Eelen 2001; Watts et al. 2005 [1992]). First-order politeness has to do with how “polite behaviour is perceived and talked about by members of socio-cultural groups” and with “commonsense notions of politeness” (Watts et al. 2005 [1992]: 3). Second-order politeness has mostly to do with theoretical concepts and generalizations, and how (im)politeness theory can explain certain language use. One application of the notion of first-order politeness concerns meta-discourse: the labels used by speakers to describe various impolite utterances, through which a researcher can access how people understand, conceptualize, and evaluate such utterances (i.e., calling an utterance rude, insulting, or disgusting). Such labels can indicate, at least potentially, hearers’ perceptions of impoliteness, as well as reveal something about the conventionality of impolite expressions (Culpeper 2011; Eelen 2001; Mills 2003).

At the same time, the discursive approach has been criticised for not contributing to a theory of (im)politeness, since it focuses excessively on actual conversations and interactants’ specific conceptualizations of (im)politeness (Terkourafi 2005a). A natural antidote to using only one approach with its inherent problems is to combine first-order and second-order (im)politeness together with ideas from Brown and Levinson’s face theory, as we do in the present study.

2.3 Conventionality and conventions

The concepts of conventionality and convention are complex and “multidimensional” (Schmid 2020: 89). A convenient starting point is the following definition, provided by Lewis (1969):

A regularity R in the behaviour of members of a population P when they are agents in a recurrent situation S is a convention if and only if, in any instance of S among members of P,

1. everyone conforms to R

2. everyone expects everyone else to conform to R;

3. everyone prefers to conform to R on condition that the others do, since S is a coordination problem and uniform conformity to R is a proper coordination equilibrium in S. (Lewis 1969: 42)

Condition (a) implies that behaviour that is conventional is common or frequent. Criteria (b) and (c), with terms such as “expect”, “prefer”, and “proper” imply normativity, even though Lewis has been criticised for downplaying this dimension (Itkonen 2008), as well as for lacking a diversified view on being a member of a certain group (Terkourafi and Kádár 2017). A third dimension of particularly semiotic conventionality (i.e., concerning expressions in a sign system such as language or gesture) that is important for our approach, but is lacking in this definition, is what we can call encodedness. Let us consider each in turn, and how they complement one another.

An expression in language or gesture that is conventional (in a smaller or larger community) need not be used literally by “everyone”, as suggested by Lewis (1969), but at least by sufficiently many speakers and sufficiently often to make it frequent. Such repeated use of an expression ensures the emergence of stable associations with particular contexts, thus making an expression more conventional. At the same time, as Schmid (2020) points out, high frequencies of some expressions (e.g., fucking) may indicate not so much a common, stable sense but many different ones, i.e., polysemy, which can make an expression more context general. Still, a methodological advantage of taking frequency as a dimension of conventionality is that it can be relatively easily operationalized with the help of corpus data (at least in the case of language).

Conventions are also about norms of how people should or should not behave; the latter will very likely be of low frequency, but that does not necessarily make it less conventional in a culture. As pointed out above, Lewis does not explicitly mention normativity in his definition of convention, but mutual expectations imply common knowledge, which on its side implies normativity, since social norms are commonly known (Itkonen 2008; Zlatev 2011). Norms of language (or other semiotic systems) can concern different levels (Zlatev and Blomberg 2019), corresponding for example to the three levels distinguished by Coseriu (1985). At the ‘universal’ level, norms such as the maxims of Grice (1975) prescribe that one should, for example, not be contradictory, obtuse, or irrelevant. At the ‘historical’ level when defining the grammatical and semantic norms of a language, in this case English, examples like (4) and (5) are proscribed as incorrect.

Finally, at the ‘situated’ level, norms concern appropriateness, and consequently they become more context-sensitive. For example, (6) may be considered either appropriate or inappropriate depending on who says this, to whom, and under what circumstances.

Irrespective of the level, norm violations imply accessibility of speaker intuitions: “spontaneous, but still conscious judgments” (Zlatev and Blomberg 2019: 83). Methodologically, such judgements can be elicited and serve as an operationalization for the normative dimension of conventionality.

The third dimension of conventionality, encodedness, can also be traced back to the seminal work of Grice (1975). To remind, Grice distinguished not only between “what is said” versus “what is implicated”, but also between conventional and conversational implicatures; the first type is based on the conventional meanings of the words (e.g., unlike the neutral conjunction and, but implicates contrast), while the second goes beyond them and requires inference and contextual knowledge. Conversational implicature was further subdivided into generalized and particularized, where the first is more context-general, whereas the latter requires detailed knowledge of the context (Grice 1989: 37–40). The works of Neo-Gricean scholars contributed to further amendments and finer distinctions. For example, Levinson (2000) proposed to see implicatures as a cline rather than distinct types, whereas Terkourafi (2015) divided generalized conversational implicatures into two types, where one concerns utterance-type meaning presumed in all contexts, and the other concerns meaning presumed in a minimal context: “extra-linguistic features that include, but are not limited to, the age, gender, and social class of the interlocutors, the relationship between them, and the setting of the exchange” (Terkourafi 2015: 15).

This provides a cline of implicatures that corresponds to degrees of conventionality, as shown in Table 2. The further to the right in the table, the higher the dimension of encodedness. Methodologically, this is commonly operationalized in terms of cancelability tests: the more encoded (semantic) the expression is, the less it is possible to cancel (or defeat) what it says or implicates (Belligh 2021; Haugh 2015). Thus, conversational implicatures (both particularized and generalized) are more obviously cancellable than conventional ones.^[2]

Table 2:

Conventionalization process (combining and adapting Culpeper [2011: 128] and Terkourafi [2005b: 211–212]).

Inferred meaning			Encoded meaning
More time/effort			Less time/effort
Non-conventionalized	Less conventionalized	More conventionalized	Fully conventionalized
Particularized conversational implicature (PCI)	Generalized conversational implicature (GCI type 1)	Generalized conversational implicature (GCI type 2)	Conventional implicature
Utterance-token meaning derived in nonce context	Utterance-type meaning derived in a minimal context	Utterance-type meaning presumed in all contexts	Coded (sentence) meaning

As indicated in the second line of Table 2, there is an expected inverse correlation between encodedness and the time/effort needed for working out the actual meaning. Expressions that require more inference should in general require more effort in order to arrive at the intended meaning (Culpeper 2011). This prediction could in principle be tested empirically using behavioural measurements such as reaction times, and one would expect to find faster reactions for more conventional/encoded expressions.

To sum up, we propose that frequency, normativity, and encodedness are three dimensions of conventionality that are independent, but also complementary. Viewing conventionality as a multidimensional concept goes well in line with the pluralistic and non-reductionist approach of cognitive semiotics. By taking all three dimensions into consideration, we should gain a better understanding of how conventionality relates to impoliteness, as we discuss below. In addition, each dimension offers alternative ways to operationalize conventionality, which our empirical study takes advantage of.

2.4 Summary and hypotheses

In sum, conventionality has become a crucial concept in the field of (im)politeness research over the last ten years (Culpeper 2011; Terkourafi 2015). However, care needs to be taken when it comes to integrating the three dimensions of conventionality: frequency, normativity, and encodedness.

Frequency is usually gauged through corpora data, but impoliteness examples are often very few or may not come from real-life exchanges (see Culpeper 2011). Further, a cross-cultural analysis of polysemiotic impoliteness requires parallel, polysemiotic corpora with sufficiently many examples and contexts from each culture. Given these challenges, we choose to address this dimension in our experiment on the basis of judgements, i.e., perceived frequency, as we show in Section 3.

Normativity has been used in the field of (im)politeness by representatives of the discursive approach, who focus on how people talk about (im)polite behaviour. However, not that many studies have implemented the analysis of meta-discourse that would tackle the normative aspect (Culpeper and Hardaker 2017). We integrated this idea by eliciting impoliteness meta-discourse.

We treat the dimension of encodedness as the end of the cline, with inferred meaning on the opposite side, and different kinds of implicatures in between, as shown in Table 2. This allows us to make a distinction between highly and less conventional expressions. The former stand close to conventional implicature, require no inference, and are not defeasible. Precisely this feature should make them appear more impolite than less conventional and cancellable expressions.

Taking into consideration this theoretical framework and the findings of previous studies, we formulated the following hypotheses, further operationalized in Section 3. Language and gesture are two distinct semiotic systems exhibiting unique features (see Table 1). Given that gestures are more closely related to physical actions and originate in bodily mimesis, we expect impolite gestures to be perceived as more impolite than spoken expressions (H1). As several (im)politeness studies show, in Russian culture positive face has more value, whereas in Swedish culture negative face seems to be dominating, thus we expect that the perceived impoliteness of utterances damaging positive face will be higher for Russian participants, whereas for Swedish participants it will be higher for utterances damaging negative face (H2). Taking into consideration Gricean theory and fine-grained distinctions made by Neo-Gricean scholars, it follows that more conventional impoliteness is less cancellable. Thus, provided that the impoliteness level is kept constant, we expect that highly conventional utterances will be perceived as more impolite than less conventional (H3). Finally, given that more conventional expressions require less inference, we expect that highly conventional impolite utterances will be evaluated faster than less conventional expressions (H4).

3.1 Materials

The stimuli consisted of short dialogues where each speaker had two turns. These dialogues were taken from The Russian National Corpus (https://ruscorpora.ru) with some of them undergoing changes in context, when one public place was substituted by another (i.e., café by library). Some dialogues have undergone changes in their final impolite expressions. For example, the original dialogue contained a word such as fool, whereas we needed a more impolite expression such as those in (7)–(10) or a gestural expression such as in (11)–(12). When we had to change impolite expressions, we relied on native speakers’ intuitions, but we also checked frequencies in corpora. With less conventionalized expressions and gesture, we mostly relied on intuitions as such expressions are not well represented in corpora.

Once a Russian version of the script was compiled, it was translated into Swedish and checked by several native speakers of Swedish. Investigating two distinct semiotic systems (i.e., language and gesture) in two distinct cultures required different approaches for establishing the comparability of spoken and gestural expressions. For verbal expressions, we considered: (a) their literal and pragmatic (contextual) meanings; (b) their potential impoliteness level; (c) their frequencies in corpora. In ideal cases, these factors would be at the same or comparable level both for Russian and Swedish. In problematic cases, we prioritized frequencies and potential impoliteness over meaning. For example, while (8) and (9) in Russian may seem possible equivalents for the highly conventional derogatory Swedish phrase (10), expression (8) does not reach the same level of impoliteness and is not fully conventional, whereas (9) includes a very strong intensifier. Hence, we adopted (7) as an appropriate match in terms of conventionality and impoliteness level. Fortunately, such cases were very rare.

Table 3 provides a sample dialogue in the two languages, with an English translation.

Gestures were matched both for their meaning and conventional articulation in each culture. For example, the “crazy” gesture was performed differently in Russian (11) and Swedish (12).

Most of the gestures had identical articulation in both cultures, such as an emblematic middle finger, index-finger pointing, and iconic gestures based on resemblance with the denoted action. To double check our intuitions about articulation, impoliteness level, and conventionality of some gestures, we conducted interviews with five native speakers in each language who agreed with our intuitions.

After the full script containing 32 dialogues with impolite expressions was compiled, we tested our intuitions regarding impoliteness level and conventionality of expressions several times. At first, we created a pilot online questionnaire that was distributed to random samples of 20 participants in each language. After this, 16 dialogues that were found problematic were removed, and the remaining 16 were complemented with 16 control dialogues (with polite or neutral utterances). They were all checked by native speakers. This final version of the questionnaire was completed by 54 Russian and 72 Swedish participants. Only at this stage did we move to recording video stimuli for the experiment, where the script from the final questionnaire served as the basis.

The experimental stimuli included 44 items per culture (16 target items, 16 control items and 12 training items). We recorded both the target and control videos, with male and female acquaintances as “actors”, while most of the training items were taken from the internet. The setting of all dialogues was either identical or as similar as possible in the two cultures. The target dialogues were equally distributed across three conditions, resulting in two items for each unique combination of categories: (a) Impoliteness level: High & Low; (b) Conventional: Highly & Less; (c) Semiotic system: Language & Gesture. The only condition that was not balanced was the type of face being damaged, with 7 out of 8 gestures damaging negative face, whereas 6 out of 8 expressions in language damaged positive face.^[3] How the dialogues were distributed across all conditions is shown in Appendix B. The full scripts in Swedish and Russian are in given in Appendices C and D, respectively.^[4]

Each dialogue was recorded as a video clip with two interacting people (actors), where at first there was a short exchange, and then actor B performed an impolite (target), or either polite or neutral (control) act towards A. This final “punchline” (see Figure 1) was always at the very end of the clip. Gestures were used in the absence of any vocalizations, and language was used in the absence of emblematic or action-like gestures such as throwing or pushing. The actors were one male dyad and one female dyad per language, with the dialogues and conditions (except for face, as noted above) being equally distributed across the dyads. The total number of polite and impolite punchlines performed by each actor was counterbalanced. The recorded videos had duration from five to 15 s. Their order was randomized for every participant automatically, as described Section 3.3.

Figure 1:

Stills corresponding to the final punchline in Dialogue 1 (see Table 3), for the Swedish and Russian videos.

3.2 Participants and equipment

Thirty Russian (15 females, mean age 24.2) and thirty Swedish (11 females, mean age 22.7) native speakers took part in the experiment and the post-experimental interview. They were all university students or had recently completed their studies. The exclusion criterion was having any other first language, except for the cases of bilingualism. Participants were recruited via personal contacts in Lund (Sweden) and in Saratov (Russia).

A full-HD camera Panasonic HC-V785 with a tripod was used to record a set of stimuli. Videos were edited in Sony Vegas Pro (version 13.0). For presenting the stimuli, randomisation, and keeping track of reaction times and judgements, the E-prime software (version 3.0) was used. The stimuli were presented either on a Samsung H850 WQHD PLS monitor (24 inches, resolution 2,560 × 1,440) or on HP EliteBook 840 G4 (14 inches, resolution 1,920 × 1,080), depending on where the experiment took place. For entering responses, participants used a computer keyboard, as described below. An audio recorder was used in all interview sessions. The subsequent statistical analysis of the collected data was done in R-studio (version 2022.07.2+576) for Windows.

3.3 Procedure

Every participant was provided with a form of informed consent and asked to read and agree on participation in the experiment and in the follow-up interview. Participants were informed that they could terminate their participation at any stage. Then, the participants reported their age, gender, and their first language(s).

After this, participants received both oral and written instructions regarding the procedure (see Appendices E and F for the instructions in Swedish and Russian, and Appendix K for translation, see footnote 4). They were told that the aim of the experiment was to know how polite or impolite the “punchlines” in the videos were, and that they would have to use the scale: very impolite, impolite, neutral, polite, very polite. They were randomly assigned to one of the two versions of the scale (from left to right or from right to left, see Figure 2), to avoid any handedness effects. Participants could ask any questions regarding the instructions or procedure before the experiment and after each training session.

Figure 2:

The two versions of the scale, to neutralise possible handedness effects (with English translations of the terms, the originals being in Russian and Swedish).

The software randomized the videos, displaying them one at a time. Prior to each video-clip, the question appeared on the screen: “how polite or impolite is the last behaviour in the following video clip?” The question was shown to participants for 3 s. After this, a fixation star was shown in the middle of the screen for 1 s, and then a video was played. After the video clip ended, “RATE!” appeared on the screen, and the participants had to enter their judgement as quickly as possible by pressing the corresponding button.

The participants completed two training sessions to familiarize themselves with the task. Once they pressed a space button, eight training videos taken from various social media were shown. After finishing the first session, they had to press space to start the second session having four videos with actors. Having completed both sessions, participants could start the real experiment containing 32 videos after pressing a space button.

After the experiment, each participant was interviewed individually. Interview questions were of three types: general (G), specific (S), and additional (A) (see Appendix G). In the beginning, we asked general questions to elicit participants’ overall impression of the experiment. Then each impolite video was played again, and a set of specific questions regarding the punchlines was asked. Question S1 served for eliciting impoliteness meta-discourse and understanding the normative aspect of conventionality. Question S2 concerned the dimension of encodedness from a first-order (im)politeness perspective (see Section 2.2), which was approximated by asking participants how “direct” or “indirect” the expressed impolite intention was. Question S3 focused on participants’ impressions of how common the utterances in question are, and thus served as an indication of the frequency dimension. A set of additional questions was asked at the very end of the interview to help understand the differences in perception of impoliteness expressed through different semiotic systems and to get an overall impression of direct and indirect impolite expressions.

The Russian participants were interviewed in Russian, while for the Swedish participants Swedish was used for eliciting impoliteness meta-discourse and English for other purposes. The whole procedure took around 60 min, of which around 20 min were spent on the reaction-time experiment and the rest on the interviews. For participation, Swedish respondents received a cinema ticket funded by Lund University, whereas Russian respondents could receive the translated results of the investigation upon request.

For data analysis, we used two different mixed-effects models in order to accommodate our data with each participant providing 32 judgements. Using generalized linear models would have violated the independence assumption and potentially led to false results, prone to Type 1 error. Using a mixed-effects model, on the contrary, accounts for the dependencies in our dataset and draws correct inferences (Winter 2020). For H1, H2, and H3 we used the logistic mixed-effects model since our variables of interest (dependent) were binary: “very impolite” or “other”. For H4 we employed a linear mixed-effects model as our dependent variable was numeric – time in milliseconds.

3.4 Operationalizing the hypotheses

The methodological design allowed us to operationalize the four hypotheses presented at the end of Section 2 as follows:

1. H1. Gestures will be rated as more impolite than spoken expressions.

2. H2. Utterances (i.e., spoken expressions or gestures) that damage positive face will receive higher impoliteness ratings by Russian participants than by Swedes; utterances that damage negative face will conversely receive higher impoliteness ratings by Swedish participants than by Russians.

3. H3. Highly conventional utterances (both spoken expressions and gestures) will be rated as more impolite than less conventional ones.

4. H4. The reaction times needed for evaluating the impoliteness of spoken expressions and gestures will be considerably higher for less conventional utterances and lower for highly conventional ones.

4.1 Impoliteness in gestures versus language

Against the prediction of H1, participants’ judgements concerning the degree of impoliteness of the utterances in the two different semiotic systems were in fact quite similar, as shown in Figure 3.

Figure 3:

Judgements provided by participants for each semiotic system.

We can notice a higher number of “impolite” ratings for gesture (219 vs. 188), as well as a slightly higher overall rate of “impolite” and “very impolite” judgements about gesture (396) than for language (378), but as the inferential statistics analysis (a logistic mixed-effects model) showed, such differences were not statistically significant (p > 0.995).

In terms of impoliteness meta-discourse, the participants provided a range of labels for both gestures and spoken expressions, as shown in Figure 4 for Russian and Figure 5 for Swedish. However, the frequencies of some items differed considerably. For example, грубо (‘rude’) was used for language more than for gestures in Russian, but the opposite was the case for otrevligt (‘unpleasant’) in Swedish. Thus, while H1 was not supported, there were some indications that gestural and spoken impolite expressions were evaluated rather differently, as we discuss in Section 5.

Figure 4:

Frequencies of some impoliteness meta-discourse labels provided by the Russian participants.

Figure 5:

Frequencies of some impoliteness meta-discourse labels provided by the Swedish participants.

4.2 Impoliteness judgements by Russian versus Swedish participants

As demonstrated in Figure 6 (showing counts after the face variable was balanced by taking out dialogues number one and four, see Appendix B), there were indeed differences in how the Russian and Swedish participants evaluated impolite utterances in the experiment, but not as predicted by H2, namely that Swedish speakers would be more sensitive to utterances damaging negative face, and Russian speakers to those violating positive face.

Figure 6:

Number of judgements about positive and negative face in Russian and Swedish.

Indeed, the Swedish participants considered damaging negative face as “very impolite” more often than positive face (99 vs. 68). Contrary to our prediction, the Russian participants showed a similar pattern (64 vs. 37). The logistic mixed-effects regression analysis showed no significant main effect of face type on “very impolite” judgements in both cultures and no significant interaction between face and culture. Thus, H2 was not supported, as operationalized. However, when we compared how often the Swedish and Russian participants evaluated the impolite utterances as “very impolite” (167 vs. 101), irrespectively of the variable face, there was a strong statistically significant difference (p < 0.001). We return to these findings in Section 5.2.

4.3 Impoliteness in highly versus less conventional expressions

As shown in Figure 7, conventionality had a strong effect on how impolite utterances (spoken and gestural) were judged to be. If we combine “very impolite” judgements of the two upper (HC) and the two lower (LC) sub-plots, we can notice a difference that is more than double (255 vs. 112). The same pattern is observed when categories having the same level of impoliteness (HI or LI) are contrasted (280 vs. 87).

Figure 7:

Number of judgements about each level of conventionality and impoliteness.

The logistic mixed-effects regression analysis showed that both the effects of conventionality (p < 0.001) and the impoliteness level (p < 0.001) were highly significant.

The interaction between the levels of conventionality and impoliteness was not significant (p > 0.125), which was not surprising, as this simply means that conventionality did not have a stronger effect on judgements of impoliteness of expressions that had high impoliteness (HI, the two left plots in Figure 7) compared to those that had low impoliteness (LI, the two right plots). The fact that it had about the same effect in both cases testifies to the robustness of the influence of conventionality.

The analysis of meta-discourse representations used in the post-experimental interviews granted further support for the hypothesis. For example, Tables 4 and 5 show some common labels used by the Russian and the Swedish participants respectively, with the majority of the labels being predominantly used to describe HC rather than LC utterances. The overall number of labels provided for the HC utterances was considerably higher than for LC (708 vs. 553), and the same tendency was observed when comparing HI and LI (787 vs. 477) categories. Thus, in sum, we may consider H3 as supported.

4.4 Conventionality and reactions times

As Table 6 shows, the participants needed considerably less time to make a judgement when both impoliteness level and conventionality were high. The mixed-effects regression analysis showed that the main effects of impoliteness (p < 0.009) and conventionality (p < 0.004), as well as their interaction (p < 0.048) were significant. The general pattern can thus be clearly interpreted as supporting H4: the participants evaluated highly conventional expressions faster than less conventional ones. Further, even though the interaction between the two predictors was found to be moderate, the results showed that the effect of conventionality on reaction times also depends on the level of impoliteness. When considering the levels of the two variables, a significant difference in reaction times was found for HCHI expressions only. However, when comparing whether all HC expressions were evaluated faster than LC, without including the interaction, the difference was still significant (p < 0.032).

5.1 Descriptions of impoliteness in gesture and language

While the results did not support the hypothesis that gestures would generally be considered as more impolite than spoken expressions, we pointed out that there were some differences in the distributions of the meta-discourse labels (see Figures 4 and 5). Further, the utterances in language and gesture were described rather differently in the interviews when the participants answered questions A1, S2, and S3 (see Section 3 and Appendix G). While answering A1, some participants, including one with a background in the police service, noted that language and gesture differed not so much in terms of impoliteness but in their “aggressiveness”, “physicality”, and their capability to progress to something physical, as shown in a comment by a Russian participant in (13) and a Swedish participant in (14).

Other participants highlighted the “irritating” and “provocative” nature of gestures, and their ability to trigger negative emotions, as expressed by a Swedish participant in (15), and a Russian in (16).

In the interviews, we addressed the encodedness dimension of conventionality by asking question S2, using labels such as “direct” and “indirect” (see Section 3.3). Interestingly, the respondents labelled gestures as “indirect” more often than expressions in language (153 vs. 115). Gestures in the low impoliteness (LI) category were referred to as “indirect” most often. The participants tended to find indirect expressions as unclear or hard to grasp, and sometimes the notion was interpreted as ambiguity or vagueness, as was pointed out by a Russian participant in (17). For expressions in language, on the other hand, there was a clear pattern with highly conventional utterances being perceived as “direct” and less conventional as “indirect”.

Thus, there were indeed differences between how language and gestures were perceived in relation to impoliteness and conventionality, though in a manner that is rather hard to pinpoint, for both the participants and for ourselves. Labels like “physicality” and “aggressiveness” can be linked with bodily mimesis (see Section 2.1), but attributions of vagueness require more analysis, as this can hardly be an attribute of all gestures (e.g., emblems).

5.2 Face, culture, and impoliteness

The hypothesis about positive/negative face differences between the cultures was not supported as operationalized. This could be due to several possible reasons. First, the distribution of stimuli across the face variable was not balanced, and there was a partial conflation of the face and semiotic system variables, as acknowledged in Section 3. However, since we found no significant differences between impoliteness in language and gesture, this factor is not likely to be decisive.

Second, the split of face into the positive and negative types has already been questioned and criticised in politeness research (Bargiela-Chiappini 2003; Meier 1995) mostly because negative face can be subsumed under the strategies that are typically associated with positive face (Meier 1995: 385). Although drawing the same conclusion concerning impoliteness would be rather simplistic, the results of our study could suggest that the applicability of this distinction to impoliteness may also be problematic. After all, saying “get out!” to someone is as indicative of not valuing this person as exerting one’s power over them. Perhaps, all impolite expressions primarily damage face in Goffman’s understanding (1967), whereas the positive and negative types can indeed be seen as “two sides of the coin” (Ide 1990: 76). One more reason might be in how much a particular culture gravitates towards a specific kind of face – a parameter that is hard to operationalize. Finally, implementing a judgement scale with three options such as low, moderate, and high could have been more beneficial for testing such a hypothesis, as it introduces more variation in responses, and thus some subtle differences might become evident. On the other hand, such a design might have an impact on reaction times, as more options may require more time for making a decision.

The most interesting finding about differences between the cultures was that the Swedish participants evaluated both gestures and spoken expressions as “very impolite” significantly more often than the Russian participants did. What could account for these differences? A possible explanation may be that impolite behaviour may in fact be more frequent in Russian than in Swedish culture, at least in the present socio-political context. By being frequently faced and used by speakers, impoliteness may have become more “naturalised” (Chandler 2017: 173) in Russian culture, and consequently its derogatory effect may have diminished. Support for this speculation can be found in the replies to question S3 concerning (perceived) frequency, where the Russian participants replied that very impolite language was “frequent” and “common” more often than the Swedish participants did.

5.3 Impoliteness and conventionality

It was a major finding that the degree of conventionality contributed to how impolite a given spoken expression or a gesture was judged, independently of the prior impoliteness level that we had assigned to the expression. But which of the three dimensions of conventionality seemed to play a larger role? Interestingly, the presumed frequency of impolite utterances did not seem to have an effect, considering answers to question S3 in the interviews. In fact, presumed high frequency seemed to be linked with lower impoliteness, at least according to some participants, as shown in (18) and (19). Note also that the latter, produced by a Russian participant, is supportive of the idea of diminished impoliteness effect due to a generally higher frequency suggestion above.

This leaves the dimensions of normativity (operationalized through meta-discourse labels) and encodedness, based both on the answers to S2 concerning “directness”, and, more importantly, on our own systematic intuitions about how cancellable the impolite interpretations of the expressions were in terms of the cline shown in Table 2.

With respect to normativity, we saw in Section 4.3 that the most negative labels like “aggressive” and “rude” were more commonly given to the utterances that we had categorized as highly conventional (see Table 6). This corroborates the effectiveness of using impoliteness meta-discourse for analysing the impoliteness level and conventionality of expressions. While analysing the labels, not only did we find a larger number of labels for HC and HI categories, but also substantial qualitative differences with “stronger” labels being used for these two categories (see Appendices H and I).

Unfortunately, we could not make any firm conclusions based on the participants’ replies to question A2, as some claimed that “indirect” expressions were to be judged as more impolite since the speaker “puts more effort into coming up with it”, whereas others found such expressions less impolite, as one could “cancel out what was said”. The latter was much more in line with how we had operationalized the dimension, i.e., in terms of degree of cancelability.

In sum, given that there was a substantial overlap between first-order impoliteness, as reflected in what participants actually said (i.e., our operationalization of normativity), and the more theoretically grounded notion of encodedness, we find both dimensions of conventionality as theoretically relevant and methodologically complementary. For example, while encodedness is key when designing the study and setting up the particular conditions, thus functioning as an independent variable, the meta-discourse labels provided by speakers are a convenient way to capture conventionality as a dependent variable.

5.4 The influence of conventionality on reaction times

The conclusion above is further supported by the final and robust result about conventionality contributing to fast judgements: both high-impolite and low-impolite utterances were judged faster when they were highly as opposed to less conventional. The straightforward interpretation is that highly conventional (i.e., encoded) expressions required less inference. The level of impoliteness also had a significant effect on reaction times, but the strongest influence was observed for highly conventional expressions with high impoliteness (HCHI). Interestingly, there was no significant difference in reaction times between spoken expressions and gestures, implying that at least some gestures may have as strongly “encoded semantics” as words. Notably, this contradicts common replies in the interviews about gestures being less frequent and more indirect (see Section 5.3), which once again questions the applicability of frequency as a relevant dimension of conventionality, at least as long as impoliteness concerned.