A corpus-assisted analysis of indexical signs for (im)politeness in Japanese apology-like behaviour

1 Overview

A large part of the literature on Japanese (im)politeness is focused on the use of honorific forms as “grammatical linguistic devices typically interpreted as markers of deference” (Pizziconi 2011: 45). However, as a number of previous studies have already pointed out (Calvetti 2020; Pizziconi 2011; Shibamoto-Smith and Cook 2011, among others), the sole use or omission of honorifics does not automatically produce polite or impolite utterances and, conversely, perceptions of (im)politeness can be achieved via non-honorific linguistic devices as well. In this paper, I try to detect some highly indexical, non-honorific expressions that, when used in co-occurrence with devices popularly (albeit non always) associated with apologies, may convey an (im)polite attitude, acting, I will argue, as indexical signs for (im)politeness. They overlap with what Calvetti (2020: 103, drawing in turn from Leech 2014: 237) defines as “trigger expressions”, namely “pragmatic actions that lead to a certain interpretation of the language message”. Other labels employed in the literature are “considerate expressions” i.e., implicit polite expressions “the content of which is considered to be polite when we judge from their situational backgrounds” (Ide 1974: 128) and “(im)politeness markers” or “politeness formulae”, i.e., “recurring linguistic elements that are stereotypically associated with the communication of polite meanings” (Schlund 2014: 1). Building on previous works on indexicality, here the label “indexical signs for (im)politeness” is preferred because it emphasises that these linguistic forms do not have precise or fixed meanings, but their pragmatic functions in practice are activated by contextual features (Eckert 2008; Ochs 1992; Pizziconi and Christie 2017). It is often the case that such contextual functions and meanings are inferred when the linguistic form is used with peripheral meanings, motivated by metaphorical extension of its prototypical ones. To provide a simple example from English, the special adverb next (to) grammaticalizes as a downtoner, as in “The altruistic pursuit is next to impossible” (COCA, MAG, 2015; Davies 2008, cited in Brinton 2021: 40). Here next to is not used with its primary proximal meaning, but it rather downgrades a direct opinion, which is a speech act potentially face-threatening for the receiver. Two additional features of indexical signs for (im)politeness are also worth mentioning. First, their meaning, though conventionalised to a certain degree, arises through inferencing, thus potentially allowing for cancellability. Second, their meanings in context are probabilistic rather than categorical because they are subjected to contextual and individual variation. In light of this, I consider as indexical signs for (im)politeness a number of morphological, syntactical and lexical elements which, when used in a specific context and with a specific co-text, acquire secondary, pragmatic meanings conventionalised (to some degree) for the expression of (im)politeness. These additional layers of meaning are not fixed and can be eventually cancelled by subsequent contextual information. In the present paper, apology-like behaviours (a terminology aimed at emphasising that the use of a verbal or non-verbal behaviour usually associated with apologies does not automatically determine the illocutionary force of the act) were chosen as the specific context for the analysis of indexical signs for (im)politeness. This choice is based on the assumption that they are not relevant in terms of exchange of information, but “their significance lies on the interpersonal level of rapport” (Coulmas 1981: 70). It therefore follows that in their surrounding co-text, producers are likely to show an (im)polite attitude toward the partner’s face, possibly conveyed through the exploitation of indexicals.

Following this introduction, Section 2 presents a brief overview of Grice’s (1975) Cooperative Principle, which will be an important starting point for the description of indexicals in interactional contexts. Section 3 outlines the bottom-up, corpus-assisted approach used to select indexicals before introducing the main research questions. Section 4 summarises previous research focussing on the morphosyntactic features of the expressions chosen for the analysis and supporting a form-to-function, discursive approach to (im)politeness. The main findings are reviewed in Section 5, and their theoretical implications are discussed in Section 6. Section 7 concludes the paper by summarising the argument and reflecting on potential avenues for future research.

2 Understanding indexicals through Grice’s Cooperative Principle

One central issue in pragmatics is how the producer generates or implies – and conversely how the receiver understands or infers – some meaning beyond the literal meaning of words. Grice’s (1975) Cooperative Principle (CP) attempts to explain how interactants get from what is said to what is meant, given that we are generally able to communicate successfully even if we often say less than what we mean. In his work, Grice’s formulates the CP as follows:

Make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged (Grice 1975: 45).

He then goes on and distinguishes four categories of maxims that are manifestations of the more general CP. Such categories are as follows:

1. Quantity: make your contribution as informative as is required;

2. Quality: try to make your contribution one that is true;

3. Relation: be relevant;

4. Manner: be perspicuous.

   (Adapted from Grice 1975: 45–47)

Since these maxims are not entirely clear-cut, when applied to spontaneous data they often overlap. Nevertheless, they offer a helpful toolkit for describing how the production and interpretation of speech acts work in spontaneous interactions. According to Grice, when someone intentionally produces an utterance which is not in accordance with the maxims, i.e., it flouts one or more maxims (something that arguably happens more often than not), they do so not because they are being uncooperative, but because they are trying to convey extra layers of meaning. I will argue that this is exactly the case of indexical signs for (im)politeness, although the degree to which deviance from the maxims is self-conscious can vary.

Importantly, the applicability of the CP depends on context. For example, a Gricean approach to the type of indexicals investigated here provides quite precise criteria for the analysis of those instances where one participant in the interaction employs a form or a strategy that does not fulfil other people’s expectations. Such unfulfilled expectations would be highly relevant for the receiver(s), as they would constitute an indication of different contextual evaluations, or even consciously aggressive linguistic behaviour. However, it may be inadequate for the analysis of the “unselfconscious compliance […] with sociocultural constraints on language usage” (Jary 1998: 7) resulting in unmarked behaviour – what has also been referred to as “anticipated politeness” (Fraser 1999, cited in Terkourafi 2001) or “politic behaviour” (Watts 2003). Indexicals that supply more information than is required, but which are nevertheless highly expected, do not appear to explicitly flout Grice’s maxims. Hence, they may require a slightly different approach, which will be addressed in Section 4.2.

3 Method and aims

As already mentioned, the present study specifically investigates Japanese indexical signs for (im)politeness employed in the immediate co-text of lexical devices conventionally associated with the speech act of apologising. In order to collect a significant number of utterances, I used as search terms for the corpus collection three Japanese expressions commonly signalled as apologetic, namely gomen ‘sorry (casual)’, su(m)imasen ‘(I’m) sorry (formal)’ and mōshiwake arimasen ‘(I’m) sorry/I apologise (very formal)’. Here the labels casual, formal and very formal are preferred to non-polite, polite and very polite (also employed in the literature) to stress the fact that they indicate the formality level of linguistic markers, rather than the markedness of the utterance in terms of perceptions of (im)politeness. It follows that formal constructions tend to index a greater social distance between interactants but are not necessarily perceived as more polite than their casual counterparts. Gomen is classified as casual because it does not present the suffix -nasai (from the honorific verb nasaru ‘to do’), as its formal counterpart, i.e., gomen-nasai, does. Sumimasen (often written with the omission of the first ‘m’ as suimasen) is categorised as formal because the verbal stem sumi- is followed by the verbal suffix -masen (the negative form of the polite -masu), which signals a certain degree of formality. Finally, mōshiwake arimasen is very formal because two different choices are interacting here: it exhibits the honorific, deferential form of apology mōshiwake, followed by the verbal stem ari- and the verbal suffix -masen (again, the negative form of the formal -masu) (Usami 2002).

The corpus was collected using the free software BootCat (see Baroni and Bernardini 2004 for more information on the BootCat toolkit) from the written Q&A Japanese internet forum Yahoo! Chiebukuro (https://chiebukuro.yahoo.co.jp/), an asynchronous environment where rights and obligations prevail, interactions are predetermined to some degree and the standard form of interaction is either formal or very formal.^[1] The corpus was then uploaded onto Sketch Engine (Kilgarriff et al. 2014) and the concordance tool was employed to download all instances of IFIDs within their wider co-text (up to 500 characters on each side of the node). However, a first exploratory analysis showed that sometimes different threads were mixed within a single entry, a factor which occasionally required the manual segmentation of data. Duplicates in the corpus were manually eliminated and an analysis was conducted on 397 occurrences of gomen, 326 expressions with su(m)imasen, and 343 instances of mōshiwake arimasen, for a total of 1,066 discourse units analysed combining quantitative overview techniques and qualitative analysis. Finally, patterns of co-occurrence that showed a shift either in the illocutionary force and/or in the (im)politeness level of the speech act were manually selected. The manual interrogation and analysis of the corpus, typical of Corpus Assisted Discourse Studies or CADS (Partington 2003; Partington et al. 2013) methodology, was made possible by the small size of the corpus. To ensure transparency, the concordance corpus on which the analysis is based can be accessed through the link in the Appendix.

Results show a total of 519 entries containing one or more of what I identified as the three most frequent indexical signs for (im)politeness, namely the low-degree adverb chotto ‘a little’ (64); the verb-ending form -te shimau (296); and, finally, the conditional clause -tara (159). The formal version of chotto, i.e., sukoshi ‘a little’, and the casual version of -te shimau, i.e., -chau, will be also touched upon (see Sections 4.1 and 4.2), but a detailed analysis will be omitted because their interactional function is statistically very marginal in my data. These expressions were chosen for three main reasons. First, they present a high rate of occurrence within the wider co-text of apology-like expressions. The identification of indexical signs for (im)politeness is frequency-based because, as Terkourafi has pointed out, an expression can be considered to be conventionalised to achieve a particular illocutionary goal only if it is used frequently enough in that context (Terkourafi 2001, 2005; Terkourafi and Kádár 2017). Second, their role in conveying (im)politeness has already been discussed in previous studies (e.g., Fujii 2018; Kondo 2016; Obana 2008; Warchał 2010; Yamaoka 2004, 2017). Third, their communicative functions and inferred meanings are occasionally discussed in the corpus by interactants themselves in metapragmatic, first-order^[2] considerations that can assist the analyst in interpreting the data.

To sum up, the general hypothesis is that chotto, -te shimau and -tara, when used in co-occurrence with apology-like expressions, acquire secondary meanings that affect the illocutionary force and/or the perceived (im)politeness level of the utterance. The main research questions (RQs) are as follows:

1. Do the selected expressions indicate a change in the function of apologetic devices?

2. Do the selected expressions trigger evaluations in terms of (im)politeness?

4 Indexical signs for (im)politeness in Japanese

By way of necessary background, the following sections offer a concise overview of the three indexical signs for (im)politeness analysed here.

5 Results

Figure 1 gives the relative distributions of the three indexicals in co-occurrence with gomen, su(m)imasen and mōshiwake arimasen, as observed in the 519 concordances analysed.

Figure 1:

Percentages of occurrence of indexical signs for (im)politeness in the wide co-text of apology-like expressions.

Importantly, the above figure represents patterns of co-occurrence as observed in the extended concordance lines (up to 500 characters of co-text on each side of the node). Given the amount of co-text taken into account, Figure 1 includes instances where the three expressions investigated convey meanings not related to (im)politeness (e.g., the use of -tara in temporal constructions) – they are, therefore, only potential indexical signs for (im)politeness. Nevertheless, it allows us to make a number of assumptions.

As illustrated in the figure, generally speaking, it is possible to state that the more formal the apologetic lexeme, the less frequent the signs under analysis are – although counterexamples are not uncommon. This pattern is particularly clear for the clause-linking morpheme -tara, which co-occurs with gomen (casual) in almost one-third of the investigated instances, but is much less common with su(m)imasen (formal) and mōshiwake arimasen (very formal) (rates of occurrence are 5.2 % and 3.8 % respectively). -Te shimau (including informal variants such as -chau) presents a similar tendency, although less marked, as its relative frequency (slightly) decreases with an increase in formality of the apology-like expression (rates of occurrences are 31.2 % for gomen, 26.7 % for su(m)imasen and 24.8 % for mōshiwake arimasen). Moreover, while reading the corpus line-by-line, it also became apparent that its function as a marker of politeness is particularly prominent in utterances with mōshiwake arimasen. This is suggested by the fact that in 32 out of 85 instances of -te shimau in combination with mōshiwake arimasen, the verb-ending form is employed immediately to the left of the apologetic item, functioning as an indexical sign for politeness – a much closer default relationship with politeness than when observed in the co-text of gomen and su(m)imasen. If we now examine chotto, this adverb displays a stronger tendency to co-occur with su(m)imasen than with gomen, but is comparatively quite uncommon with mōshiwake arimasen. It should be noted at this point that, if we look at the extended co-text of apology-like expressions, its formal counterpart sukoshi (not illustrated in Figure 1) appears to be employed almost as often as chotto. For instance, when we look at the whole entry, sukoshi occurs with gomen almost as many times as chotto (22 against 24), and with a total of 22 occurrences it is also quite frequent with sumimasen (against the 34 instances of chotto). As for the extended concordance lines of mōshiwake arimasen, sukoshi is actually more frequent than chotto, with 13 occurrences compared with 6, in line with the assumption that mōshiwake arimasen requires a more formal register. While it may appear that sukoshi is as relevant as, if not more relevant than, chotto, a closer look at the data shows, first, that it is predominantly used with its literal meaning and, second, that it is never employed in the apologetic clause itself. In other words, it is not as pragmatically relevant as chotto and its connection with apology-like expressions is much weaker – hence, it will not be further discussed. These findings suggest that the combination of corpus linguistic tools with non-automatic means of annotation may be the most suitable methodology for investigating degrees of conventionalisation (hence unmarkedness) which are extremely relevant for pragmatic research.

In light of the above considerations, and in the attempt to optimise precision (Jucker et al. 2008), a second corpus query narrowing the analysis to a limited amount of co-text was carried out. Figure 2 illustrates the results of this second corpus query in terms of relative distributions of chotto, -te shimau and -tara located exclusively in the first position to the left (L1) of apology-like expressions. An exception was made for instances where -te shimau is employed in the second position to the left (L2) when the L1 position is occupied by intensifying adverbs, usually in the following construction:

Figure 2:

Percentages of occurrence of indexical signs for (im)politeness to the left of apology-like expressions.

Following Olshtain (1989: 158), I consider intensifiers as a routinised form of intensification that occurs internally to the apologetic device. In line with this, I take the collocation ‘intensifying adverb’ + ‘apology-like expression’ to be a single unit. Although less statistically relevant for obvious reasons, narrowing the analysis to specific positions in the text is in line with the assumption that additional functions tend to be associated with constrained positions (see Section 4.2). Methodologically, it also proffers clues to patterns which seem to be particularly salient in pragmatic terms, hence offering a buffer against the above-mentioned problem of precision.

The following sections integrate the quantitative findings illustrated in Figure 2 with a qualitative type of analysis.

6 Discussion and theoretical implications

In the previous sections I showed that the three indexical signs selected for the analysis may guide the interpretation of the utterance in different directions. In what follows, I discuss a few theoretical implications of the findings and I take a closer look at the link between indexicals, (im)politeness and (un)markedness.

My data suggest that the various meanings activated by chotto, -te shimau and -tara preceding apology-like expressions can be placed on a continuum which covers “slippery” apologies, more serious and apparently sincere (thus, all things being equal, more polite [Haugh and Melody-Chang 2019]) ones, as well as unmarked forms. The inclusion into the analysis of potentially negatively evaluated and unmarked forms of behaviours is reminiscent of the notion of relational work (Locher 2004, 2006; Locher and Watts 2005; Watts 2003), which encompasses all kinds of facework. A visual representation of this approach to (im)politeness is proposed by Watts (2005], [xliii) and sees politeness, unmarked or politic behaviour, and impoliteness (which includes over-politeness) as contiguous aspects of the spectrum of relational work. A tentative representation of where the patternings investigated here may be located within this spectrum is illustrated in Figure 3.

Figure 3:

Apology-like behaviours along the spectrum of relational work (adapted from Watts 2005: xliii).

The figure illustrates two arrows along which we have different sections corresponding to positively (upper half) and negatively (lower half) evaluated behaviours. In the attempt to clarify the role of indexicals in guiding interactants’ interpretations of apology-like behaviours, I have located along the figure the prototypical patternings (in italic) investigated in the previous sections, together with their pragmatic functions (in bold uppercase). At the top right, we have apologies that are more likely to be interpreted as polite because they provide a number of additional linguistic elements that conventionally increase the perceived sincerity of the utterance. In speech act terms, sincere apologies satisfy (at least partly) the conditions required for a speech act to be, in Austin’s (1962) terms, felicitous. However, the same linguistic forms can also be closer to non-polite or unmarked, rather than polite, behaviour because, in certain circumstances, they may simply “do the job” and meet interactants’ expectations without producing any inferences in terms of (im)politeness – hence passing unnoticed. On the opposite side of the spectrum, we have “apologies” (quotes intentional) that have been downgraded to varying degrees. For instance, we have seen that the close-linking morpheme -tara immediately preceding an apology-like expression may imply some degree of vagueness on whether the event in the conditional clause has taken place, or who is to blame for it. I argued that, by potentially decreasing the producer’s accountability for the meanings an apology is usually associated with, -tara potentially opens the ground for a negative evaluation in terms of impoliteness – although the interpretation remains in the hands of the receiver. Forms of over-politeness where the politeness level is higher than expected can also be negatively evaluated and are thus located in the lower half of the figure. Notably, in my data, the same construction ‘-te shimau’ + ‘apology-like expression’ appears both as appropriate polite behaviour and as over-polite inappropriate behaviour, the latter being negatively marked (see Section 5.2). This shows once again that there is no intrinsically polite or impolite speech act and socially produced patterns of expectations are key in discriminating between degrees of (im)politeness/(un)markedness. Finally, the use of apology-like expressions as attention-getters and pre-requests is often signalled by chotto in the L1 position. In the figure, this patterning is placed within the share of unmarked behaviour, but closer to impoliteness, as it can be thought of as a form of mild apology that can either pass unnoticed, or be considered as (probably slightly) impolite, since it is part of a mild threat to the receiver’s negative face (Brown and Levinson 1987).

A note of caution is in order. Figure 3 is necessarily a simplification for illustrative purposes, and it should not be taken as an attempt to suggest that two-word collocations determine, on their own, pragmatic meanings. As I hope to have shown, there is no hard line between positively and negatively evaluated apology-like behaviours, but rather a polite-unmarked-impolite continuum along which apology-like forms are located depending on the linguistic co-text, the interactional context, socio-cultural norms and values, personal preferences, etc. In line with this approach, Figure 3 is just a tentative representation based on my data.

7 Conclusions

The present paper focused on relatively small units of meaning that seem to trigger relatively consistent interactional effects. These are constructions where an apology-like expression is immediately preceded by one of the three indexical signs for (im)politeness selected for the analysis, namely chotto, a conditional clause with -tara and various forms of verb-ending construction -te shimau. The analysis was data-driven: a data set collected from the Q&A website Yahoo! Chiebukuro suggested that these patternings may be frequently associated with indirect, secondary meanings, potentially triggering (im)polite inferences. These claims were verified through different forms of triangulation: metapragmatic comments made by members of the Yahoo! Chiebukuro community, previous studies looking at both Japanese and English, and corpus queries investigating these constructions in a larger, more balanced web corpus.

As regards RQ1 “Do the selected expressions indicate a change in the function of apologetic devices?”, results show that chotto immediately preceding gomen or su(m)imasen signals that the utterance is most likely to function as an attention-getting device or as a pre-request. On the other hand, -te shimau and -tara do not appear to modify the prototypical interpretation of the search terms as apologetic devices, but they do modify their degree of perceived (im)politeness – if contextual features allow for such interpretation. In fact, and with reference to RQ2 “Do the selected expressions trigger evaluations in terms of (im)politeness?”, the findings suggest that chotto and -te shimau, generally speaking, increase the (perceived) politeness level of the utterance by mitigating a potential face-threat and expressing formulaic regret respectively – but counterexamples were also considered. In contrast, a conditional clause in combination with an apology-like expression may acquire a derogatory value: in specific contexts, it may imply that the apology remains a surface realisation of politeness and that its use is at least partly motivated by self-interest, namely the desire to limit the self-deprecation involved in the speech act of apologising. These different types of apology-like behaviours were tentatively shown in Figure 3, which is an attempt at visualising indexical meanings along an (im)politeness continuum. My findings support Grice’s assumption that, with the supposition that the overall Cooperative Principle is being observed, linguistic devices that appear to flout the maxims are in fact signalling that there are additional layers of meaning the producer wants the receiver to infer. However, the level of (un)markedness of the utterance also contributes to its evaluation, and a more refined model of indexicality needs to take into account the notion of politic behaviour.

In conclusion, we can reckon that apparently polite forms perform a number of different functions in discourse, from affecting the illocutionary force of the utterance, to conventionally expressing politeness, to conveying a potential face-threat. Considering the limited number of instances analysed here and the variety of factors that affect probabilistic correlations of form and pragmatic meaning, the findings reported above should be corroborated (or falsified) by further research. Future studies may also investigate how indexical signs for (im)politeness and their multiple meanings are related to expressions of other speech acts (e.g., criticism, promises, thanks, requests, and so on), both intra- and inter-linguistically. Describing indexical signs for (im)politeness is a notoriously slippery business and the present study is far from exhaustive. Nonetheless, it illustrates a data-driven and replicable process for functional analysis that successfully merges corpus and pragmatic approaches. This will allow for further development within (im)politeness studies in particular.