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

: This study provides a corpus-assisted pragmatic investigation of three Japanese expressions: the adverb chotto ‘ a little ’ , the verb-ending form -te shimau , conveying (formulaic) regret, and the conditional clause with -tara . These are deictic forms I refer to as indexical signs for (im)politeness because they can, under certain circumstances, trigger evaluations in terms of (im)politeness, potentially favouring an indirect interpretation of the utterance. They are investigated in co-occurrence with apology-like behaviour based on the assumption that, in this context, interactants are more likely to exploit linguistic strategies for conveying additional layers of pragmatic meaning. The main ﬁ ndings point to a wide range of possible interactional meanings the selected forms can acquire in naturally occurring data, from a ﬀ ecting the illocutionary force of the utterance, to conventionally matching interactants ’ expectations, to conveying a potentially face-threatening act. These results support the assumption that seemingly polite speech acts will not necessarily be doing polite work (or not only) and highlight the relevance of the interactional context for retrieving communicative meanings.


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 facethreatening 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 impliesand conversely how the receiver understands or inferssome meaning beyond the literal meaning of words.Grice's (1975)  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 behaviourwhat 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.

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(Terkourafi , 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 2004Yamaoka , 2017)).Third, their communicative functions and inferred meanings are occasionally discussed in the corpus by interactants themselves in metapragmatic, first-order2 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: (RQ1) Do the selected expressions indicate a change in the function of apologetic devices?(RQ2) 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.

Chotto
The first expression chosen for the study is the informal, low-degree adverb chotto 'a little'.As already mentioned, it has a formal counterpart, namely sukoshi, which occurs 57 times in my data.Sukoshi shares with chotto the same propositional content but is usually employed in higher registers, indexing a greater distance between interactants.Sawada (2015) also points out that, while chotto can express both a quantitative and an expressive meaning (i.e., attenuation in degree in the force of the speech act), sukoshi can be combined only with gradable predicates to convey scalar meanings, as in: (a) According to Sawada (2015), here chotto can function as a pre-modifier of both gradable (a) and non-gradable (b) forms, while sukoshi, is appropriate only in the former case.One explanation for this is that chotto, as used in example (b), does not quantify degree or amount, but "qualif[ies] the more abstract entity that is the speech act" (in this case, a request), whilst at the same time "minimi[sing] the rudeness or abruptness that may be implied" (Matsumoto 2001: 1).These findings are supported by Yamaoka (2004), who demonstrates that chotto is often used without its referential meaning when (c) criticising, or (d) rejecting a request to minimise the facethreat: Tōkyō sodachi no kimi niwa chotto wakaranai kamoshirenai keredo Tōkyō grew.upyou CHOTTO understand-NEG maybe but 'You grew up in Tōkyō, maybe you won't CHOTTO-understand…' (d) Sono kingaku wa chotto muri ka to omoimasu ga That money TOP CHOTTO impossible Q Q T maybe but 'I think that that amount of money may be CHOTTO-impossible…' (Adapted from Yamaoka 2004: 24) In (c) the producer threatens the receiver's positive face by negatively evaluating their self-image.However, they do so indirectly by means of lexical and grammatical devices that modify the force of the speech act.These are the low-degree adverb chotto 'a little'; the expression kamoshirenai 'maybe', conveying possibility; and the adversative conjunction keredo 'but', which, when used at the end of the utterance, involves some degree of indeterminacy and, again, hedges illocutionary force.The chotto in example (d) is immediately followed by the non-scalar adjective muri 'impossible', a construction that has no direct translation in English (Matsumoto 2001: 5) but is functionally similar to the next to impossible we have seen in Section 1.In fact, chotto here has the function to convey uncertainty and unwillingness to utter a statement that may be unwelcome to the receiver.The polite implication that the statement may be true only in certain respects is then reinforced by the hedging expression to omoimasu 'I think', followed by the adversative conjunction ga.In both examples chotto is not relevant in terms of exchange of informationthe denotational meaning of both utterances would be the same without it.With the supposition that the producer is observing the overall Cooperative Principle (Grice 1975; see Section 2), we can safely assume that the reason for the use of chotto is here on the level of affective face: by exploiting the use of chotto as downgrader and other communicative strategies to avoid a direct opinion (c) or refusal (d), the producer tries to counteract the potential face damage of criticism and rejection respectively.These usages of chotto exemplify how a lexical item can shift from its prototypical (referential) meaning to a more peripheral (more abstract) meaning with the aim of showing consideration for the feelings of others (or pretending to do so), whilst projecting a positive affective image of the producer.

-Te shimau
The second expression under analysis, also multi-faceted, is the verb-ending form -te shimau.-Te shimau has a number of possible variants depending on the verb it is attached to and its level of formality (for instance, one of its informal variants is -chau).For reasons of brevity, in what follows I collectively refer to -te shimau and its variants simply as -te shimau, also in light of the fact that in the data -chau and similar variants are quite infrequent (only 17 occurrences, of which 15 in correlation with gomen) and were never observed in sentence positions that, as will be seen, are particularly relevant from a functional perspective.This construction is grammaticalized from the verb shimau 'put (something) away/finish', which has the following three main forms: Their normal, non-grammaticalized meanings are (e) 'to put (something) away' and (f) 'to finish' (though this latter use seems to be obsolete in present-day Japanese): (e) Monooki ni yomanai hon o shimatta store.room in read-NEG book OBJ SHIMAU-PST 'I put the books that I don't read away in the storeroom' (Ono 1992: 369 Mise o shimau koto ni shita store OBJ SHIMAU-NPST decide-PST 'I decided to close the store' (Genius English Japanese dictionary, 3rd ed.) According to Ono (1992: 370), the development of shimau into the grammaticalized form -te shimau (and its variants, including also shortened versions such as -chau) can be found in the construction known as clause-linking, where the verbal suffix -te is added to a verb in the infinite form to group two clauses together.The following example illustrates the initial phase of the grammaticalization of -te shimau: (g)

Sono hon
o yonde shimatta that book OBJ read-TE SHIMAU-PST 'I read the book and put it away' (h) Sono hon o yonde shimatta that book OBJ read-TE SHIMAU-PST 'I finished reading the book' (Ono 1992: 371) In the development of additional functions, the verb shimau changes from free to constrained position and shifts from concrete to abstract meaningstwo features characteristic of grammaticalization in English as well (Brinton 2021).Moreover, the verbal suffix -te loses its connective function, as supported by the fact that the utterance in (h) gets to encode a single event.-Te shimau as a marker of completion further developed to convey a modal, pragmatic meaning (i): 'I was scolded-TE SHIMAU by the teacher' (Tamura 2007: 11) This latter use of -te shimau is of particular interest for the present study because it expresses the producer's negative feelings and attitude toward the event described in the clause -te shimau is attached to.This is sometimes referred to as "boulomaic modality" or "emotional attitude", which "concerns an indication of the degree of the speaker's (or someone else's) liking or disliking (affectively) of the state of affairs," (Nyuts 2018: 39).More specifically, -te shimau conveys the producer's disappointment and regret towards an event seen as regrettable and which they assume they have no control over (Fauconnier 2013: 37, 38).As will be seen in Section 5.2, the verb-ending form -te shimau co-occurring with apologetic items exploits this latter pragmatic function to address the receiver's positive face and mitigate a potential threat to face.

-Tara
The present study also investigates indexicals whose use is not characterizable in terms of mitigation because they may signal a not-so-polite attitude.I will argue that this is the case of the conditional clause with -tara.The Japanese language presents a number of different clause-linking morphemes that can be used to convey what in English is referred to as a conditional clause, and that, by and large, can be translated in English with the if-clause.Here I do not venture into an extensive description of such constructions, but focus specifically on -tara, as this is the form that seems to be primed to co-occur with apology-like expressions.
Morphologically, the -tara construction involves the past tense of the predicate, hence it can appear in different forms depending on the lexeme it is attached to.When employed as an inflectional ending of verbs, it can take the form -tara, as in tabetakat-tara 'eat-COND', or its allomorph -dara, as in non-dara 'drink-COND'.When attached to adjectives, it usually takes the form -tara, as in samukat-tara 'cold-COND'.Finally, dat-tara is the inflected form of the copula da (Fuji 2018; Han 1966).These different constructions are summarised here as conditional clause with -tara.Functionally, -tara is polysemous in nature and can be located in the intersection between conditional and temporal clausea variation which is by no means unique to Japanese.These two possible meanings are exemplified below: Sōki ni shujitsu o shi-tara tabun yoku naru.early at operation OBJ do.COND perhaps well become 'If (she) has an operation at an early stage, she will perhaps recover.' (Adapted from Fujii 2018: 562 Nihon e it-tara denwa shimasu.Japan to go-COND call 'If/when (I) come to Japan, I will call you.' (Adapted from Fujii 2018: 563) While (j) is prototypically conditional, in that it conveys the producer's assumption about a causal relation between the predicate and the main clause, depending on the context (k) can convey a meaning that is normally regarded as temporal (Fujii 2018: 562, 563).My working hypothesis is that the conditional (and not the temporal) clause with -tara in apologetic contexts illustrates an additional layer of pragmatic meaning the producer can strategically exploit for their own interactional goals.Hence, it is of fundamental importance to distinguish between conditional and temporal constructions with -tara.This requires some degree of manual (and often laborious) annotation of the data, because "form-function mismatch of most pragmatic phenomena means that automative assignment of tags will often lack precision" (O'Keeffe 2018: 599).This motivates the corpus-assisted methodology employed here.

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.
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 arealthough 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 politenessa 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 weakerhence, 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: V-te shimai/-te shimatte taihen mōshiwake arimasen V-TE SHIMAU really sorry 'I am really sorry for V-TE SHIMAU' 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.

Chotto as a marker of illocutionary force
In my data, chotto occurs relatively significantly in the L1 position of su(m)imasen (rates of occurrence 2.5 %), whereas it was found only in 0.5 % of apology-like clauses with gomen.Notably, it was never used as a collocate of mōshiwake arimasen.This may be due to the difference in degrees of formality chotto and mōshiwake arimasen exhibit: while the former is non-formal, the latter is the deferential form of apology.When used in the L1 position of gomen and su(m)imasen, chotto may serve as a hedge of the apology, as in: Arara chotto gomen na, an chan.Soko, doite kureru ka.Soitsu Ehy CHOTTO sorry SFP young man There move BEN Q He wa ore no tomodachi nan da yo.
TOP my friend N be SFP 'Ehy, CHOTTO-sorry, young man.Would you move away from there?He is a friend of mine.' The set chotto gomen in (1a) is a mild apology which does not remedy a past offence, but rather functions as an attention-getting device.Here chotto is not relevant in terms of exchange of informationit does not affect the denotational meaning of the utterance.These findings echo those of Matsumoto (2001) and Yamaoka (2004) (see Section 4.1) and closely align with Brown's remark that the particle ala in Tzeltal, which is usually glossed as 'a little', "appears repeatedly in conversation […] where little or no new information is being conveyed" (1980: 125).Chotto then, just like ala, appears to break Grice's maxims of Quantity and Relation (Grice 1975).However, with the supposition that the producer is observing the overall Cooperative Principle (Grice 1975), we can safely assume that the point of the use of chotto here is on the level of affective face: it signals that additional, polite meaning is implied, and that such additional meaning is to be recognised as intended by the receiver.The derived use of gomen as an attention-getter in (1a) is also observed by Coulmas, who argues that "to intrude upon someone is often regarded as an object of regret" (1981: 76) and, as such, is a behaviour that requires an anticipatory (ex ante) quasi-apology.In my data, the set 'chotto' + 'apology-like expression' is the most typical form uttered when the producer wants to catch the receiver's attention, in a way similar to the English excuse me (Deutschmann 2003: 74).Notably, in (1a) the set chotto gomen also opens the ground for a direct request, showing that a single apology-like utterance may have multiple functions at the same time.
The aforementioned assumptions on the use of chotto are reflected in the following excerpt, which reports one metapragmatic comment retrieved from the corpus.The comment is made by a member of the Yahoo!Chiebukuro's community and concerns specifically the illocutionary force of the collocation chotto sumimasen: 'Sumimasen ga ○ ○ shite kudasai', 'Chotto sumimasen' nado, irai sorry but do HON CHOTTO sorry etc. request ya yobikakeno sai ni mochiiru 'suimasen' wa, karui shazai no and call GEN when use sorry TOP mild apology GEN imi kara to kangaeraremasu.meaning from QT suppose-POT 'We can suppose that the word "sumimasen" used in requests and to get someone's attention [as in] "Sumimasen but please do ○ ○", "CHOTTO-sumimasen" etc. derives from a mild apology.' Here the producer explicitly links the use of sumimasen in requests and as an attention-getting device to its primary apologetic meaning, suggesting that mild apologies, such as chotto sumimasen, can be expected to convey additional meanings.We can interpret this as supporting the two assumptions made abovenamely, that apology-like expressions may acquire additional meanings when used in cooccurrence with chotto, and that these peripherical meanings are somehow linked to the prototypical apologetic one but carry additional information.
Since chotto is the least frequent of the three indexical signs for (im)politeness analysed here, and low frequency may skew the findings, a further step was taken.In order to test the assumption that the collocation 'chotto' + 'apology-like expression' can be used as an attention-getting device and as a pre-request, I have looked at the first 50 occurrences of chotto su(m)imasen in the web corpus JaTenTen11 (10+ billion tokens) provided by Sketch Engine.The extended concordance lines can be accessed through the link provided in the Appendix.The analysis was limited to sumimasen for space concerns, but it could be easily extended to other collocations.Among the 50 concordances analysed, 20 were found to be employed as attention-getters and 10 as pre-requests.QT asked 'A member of the staff asked me "CHOTTO sorry, but could you please call the store manager?"'A note of caution is in order.Even if a number of examples supporting the hypothesis that chotto sumimasen is employed as an attention-getter, as in (1c), or pre-request, as in (1d), were indeed observed, counterexamples were also found.Moreover, as a further analysis of sumimasen and chotto as separate search terms made clear, the use of chotto as a pre-modifier of sumimasen is statistically very marginal.However, the results above can still be significant if we are interested in how (rather than how much) interactants exploit indexicals for their interpersonal goals.

-Te shimau and the case of over-politeness
The verb-ending form -te shimau too is employed in the data to increase the politeness level of the utterance.More specifically, it was found to convey the producer's negative feelings and attitude toward the event described in the main clauseusually an event somehow related to the producer and considered to be unwanted by the receiver, thus requiring an apology.-Te shimau used in co-occurrence with apology-like expressions increases the illocutionary force of the apology, as illustrated by the following example: Osoku made mēru shite shimatte gomen ne.late until mail do-TE SHIMAU sorry SFP 'Sorry for emailing you at such a late hour.' Indexical signs for (im)politeness -Te shimau as a trigger for a polite interpretation of the message often occupies the first position to the left of the search term, as in (2a).However, it might also be worth mentioning more complex apologies such as the following: In (2b) the producer first apologises for the long message, then offers additional detail on the situationan interactional move functioning also as an explanation for the violation.These mitigating devices open the ground for a direct request.-Te shimau is repeated twice (first at the left of the apologetic device, then again before the request) and the perceived politeness level of the utterance is further increased by a rather formal registeras exemplified, for instance, by the use of the formal kochira 'here' instead of the unmarked koko 'here'.In (2c) the producer explicitly communicates their responsibility for the violation.Although we have two intervening expressions between -te shimau and sumimasen (namely, the verbal construction hansei suru 'regret', followed by the verbal auxiliary orimasu, and the intensifier hontō ni 'really'), we can observe a number of relevant features.First, it seems to me that, once again, the use of -te shimau further increases the (perceived) sincerity level of the utterance, hence conveying regret.Second, by expressing the meaning of doing something inadvertently, it also suggests that no harm was intended, while projecting a positive affective image of the producer.The interpretation of the utterance as polite is further supported by the use of the two additional devices already mentioned above, namely the humble form orimasu (instead of the unmarked imasu) and the intensifying adverb hontō ni 'really'.
I want to note that these uses of -te shimau (and to some extent also the lowdegree adverb chotto) in co-occurrence with apology-like expressions may be closer to Watts' ( 2003) "politic behaviour", rather than to traditional notions of politeness as face-enhancing "extra" work.Watts describes politic behaviour as "that behaviour, linguistic and non-linguistic, which the participants construct as being appropriate to the ongoing social interaction " (2003: 20).In other words, it is unmarked behaviour that, as such, is not critically dependent on inference and is likely to go unchallenged.Although it may not be explicitly evaluated as polite by the participants in the interaction, its absence would most likely be perceived as impolite, because it is a highly conventionalised (hence expected) rapport-enhancing device contributing towards the overall facework of the interaction.These considerations support the claim made in Section 2 that a classic Gricean approach may neglect the role conventionalisation plays in the use of unnoticed and unmarked polite formulae.
Interestingly, however, one case in the corpus revealed a pattern which could be viewed as inconsistent with the positive evaluation associated with the unmarked usage of -te shimau.In this specific example, the use of -te shimau and other linguistic devices typical of formal registers results in a politeness level higher than expected, eventually favouring a negative evaluation.In the excerpt in question, the producer reports a (failed) attempt to apologise to a friend for going back from work without waiting for them.It reads as follows: (2d) […] itsumo wa isshō ni kaeru noni hitori de kaerō to shita koto desu.Futsū, 'Gomen I'm a very petty-minded person.We won't be friends again'.' A divergence in the presumed moral order (i.e., the "seen but unnoticed" set of expectancies that regulate social interactions [Garfinkel 1964]) of interactants is at play here.The use of a politeness level higher than expected gives rise to a relative mismatch, which pertains to "the (mis)matching of (im)politeness associated with surface forms relative to the norms of the specific context" (Culpeper and Tantucci 2021: 162).In other words, an asymmetrical power relationship between the producer and their receiver may have justified, or even required, a context-appropriate mismatch in the use of polite formulae, which are indeed perfectly acceptable in hierarchical contexts.However, since the interactants are framed as friends, this is not the case.As a consequence, the use of a quite formal speech level, instead of triggering the politeness implicatures envisioned by the producer, has the opposite effect because is perceived as inappropriate by the receiver.This is in line with Watts' observation on what he terms "over-politeness": [Marked behaviour] may be perceived as negative either if it is open to an interpretation as impolite (or as downright rude), or if it is perceived as over-polite, i.e. those kinds of negatively marked non-politic [i.e., marked] behaviour tend towards similar kinds of affective reaction on the part of co-participants.Certain speakers consistently evaluate polite behaviour as unnecessary and offensive.… the communicative effects of over-polite behaviour may seem remarkably similar to those of downright rude behaviour.(2005: xliv) In (2d), the negative evaluation that follows a sequence of over-polite apologies has relevant consequences for the relationship, possibly leading to its end ("we won't be friends again").We can thus infer that extralinguistic factors, e.g., personal preferences and relative distance between interactants, can change the interpretation of otherwise polite linguistic constructions, potentially producing unintended perlocutionary effects.This supports the assumption made in Section 1 that -te shimau is an indexical device whose meaning is strongly context-sensitive and, as such, it cannot be automatically construed as polite without taking into account a number of co(n)textual factors.Among these, an important variable I did not account for is the verb -te shimau is attached to, which obviously affects interactants' interpretations of (im)politeness.More research on these co-selection processes is needed.

-Tara and those not-so-polite apologies
A number of examples retrieved from the data seem to suggest that a conditional clause with -tara, when used in the L1 position, may increase the indirectness level of the utterance, introducing the possibility of deniability.Although indirectness has often been associated with politeness (e.g., Brown and Levinson 1987), it is now commonly accepted that politeness may not be its main motivation.For instance, Terkourafi (2011) points out that indirect speech may signal that the producer is being, at least partly, uncooperative, while still leaving room for strategically denying their uncooperativeness at a later stage in the interaction.Along these lines, I will argue that a conditional clause immediately preceding an apology-like expression creates a space of opportunity for the producer to subtly suggest that they are not, or not entirely, responsible for the violation (or even that no violation has taken place), while not explicitly refusing to admit responsibility.It follows that, unlike chotto and -te shimau, the conditional clause seems to decrease the politeness level of the speech act, overlapping with strategies minimising the (producer's responsibility for the) violation.For instance, the following example was found in the data: (3a) Dareka wakaru hito, naosu hōhō ga wakaru hito someone understand person fix method SBJ know person i-tara tasukete kudasai!!! Go-meiwaku deshi-tara sumimasen.there.is-CONDsave.meHON-bother COP-CON sorry 'If there is anyone who understands, or who knows how to fix (this), please save me!!! I'm sorry if it's a bother.' The relevance of the additional information conveyed by the second conditional clause ("if it's a bother") flouts Grice's maxim of Manner (avoid ambiguity, be brief) (1975: 45-46), introducing the possibility that it may in fact not be a bother.A second example is as follows: (3b) Goji nado arimashi-tara mōshiwake arimasen.writing.mistakesetc.
there.is-COND sorry 'If there are any mistakes in (my) writing, I apologise.' Here again the conditional clause increases the vagueness and ambiguity of the utterance, potentially indicating reluctance in performing a speech act that addresses the receiver's face needs.This, in turn, may decrease the degree of perceived politeness associated with the apology (Kampf 2009: 14;Vollmer and Olshtain 1989: 213).In speech act terms, the conditional clause softens the apology because it questions the preparatory condition (Searle 1969)namely, that the producer believes that they potentially damaged the receiver by bothering them, as in (3a), or by causing them some form of inconvenience, as in (3b).In extreme cases, it can also potentially shift part of the responsibility for the violation to the receiver by implying that rudeness/impoliteness resides solely in their ears (or eyes)in other words, that they are "making it up".
But why should someone opt for an apology that may be open to a negative evaluation in terms of impoliteness?These linguistic choices, and the pragmatic meanings associated with them, are motivated, I believe, by the desire to save one own's facea factor occasionally overlooked in the literature on speech acts.Notably, this self-face-saving use of conditional clauses is by no means unique to Japanese.For instance, in his investigation of apologies in British English, Deutschmann (2003: 56) mentions that "the use of an if-clause in combination with an apology (I'm sorry if you didn't like it, for example), is worth noting.In such instances the offencive nature of the act, and/or the victim's right to feel offended, are questioned in the apology." Deutschmann (2003: 56) goes on to note, however, that this use was extremely uncommon in the spoken part of the British National Corpus, with only 9 instances observed.As Figure 2 makes apparent, this is not quite the case in my dataset.This finding suggests that, at least with reference to the context at hand and when the situation allows for such interpretation, the use of an apology-like expression in combination with a conditional clause may be relatively more conventionalised in (online written) Japanese than it is in (spoken) British English, thus potentially unmarked and closer to politic behaviour rather than impolitenessa point already mentioned with reference to -te shimau.However, an important specification is in order.In determining the degree of (un)markedness and (im) politeness of an apology-like speech act preceded by a conditional clause, a key factor is whether the event in the conditional clause involves the producer or the receiver.In Leech's terms (2014: 121), in fact, apologies are usually speaker-oriented (or S-oriented, as opposed to other or O-oriented) speech acts.However, "Sand O-orientation can be manipulated in the way a speech event is presented" (Leech 2014: 121).This switch applies also to apology-like speech acts with a conditional clause, as in: I'm sorry if I woke you up [BNC] I'm sorry if you feel I've been unfair to you [BNC] The main difference between these two utterances is who is being blamed: while in the first one the producer is still blaming the self, the second utterance places the blame on the other, thus shifting the responsibility on the receiver (Chaudhry and Loewenstein 2019: 321;Kampf 2009: 14) Whilst (3c) is, in my view, a form of politic behaviour, the conditional clause in (3d), similar to the one in the second example in English reported earlier, results in a somewhat "slippery" apology, more likely to be perceived as closer to impoliteness rather than politic behaviour.A third option is to avoid any references to the subject of the clause using a strategy of impersonality (Kampf 2009: 24-25: Leech 2014: 121), as in: Here the object of regret refers to the state of things and neither the producer nor the receiver of the speech act is mentioned in the conditional clause.This was usually the case in my corpus (see [3a] and [3b] above).It seems to me that, although downgraded to some degree, if the act is either oriented to the producer or to the state of things, it can be said to be closer to unmarked behaviour rather than impolitenessbut contextual information needs to be taken into account.The framing of this patterning as potentially not salient is supported by the fact that, at least in my data, it is rarely acknowledged in whatever follows in the thread.On the other hand, when the act blames the other, the preparatory condition is waved to the point that impoliteness implicatures are more likely to be inferred.How these form-function relationships manifest themselves in larger and more varied corpora of Japanese is an important question that awaits further research.

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(Locher , 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.
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)politenesshence 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 impolitenessalthough 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 apologylike 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.

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)politenessif 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 respectivelybut 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.

Figure 1 :
Figure 1: Percentages of occurrence of indexical signs for (im)politeness in the wide co-text of apologylike expressions.

Figure 2 :
Figure 2: Percentages of occurrence of indexical signs for (im)politeness to the left of apology-like expressions.
The most convincing examples are as follows: sorry for the long message.Since the ambiguity in my heart does not clear up (and) I don't know what to do, please let me ask for advice here.' to teineina shazaibun o nando mo okurimashita.Kekka, tomodachi kara wa 'Sasaina koto de konna teineina shazai o senaikan hodo watashi wa chīsai ningen nan ya ne' 'Mō tomodachi ni wa modoren' to iwaremashita.'[…] even if we always go back from work together, I went back on my own.It is something that usually can be remedied with one word 'Sorry [CASUAL]'.However, I thought I upset them and I sent a message with the polite apology 'I'm sorry [FORMAL] for going back earlier and forgetting-TE SHIMAU the bag'.My friend replied with 'This is not how you talk with a friend'.I didn't understand those words and I sent other many polite apologetic messages [such as] 'I'm sorry [FORMAL]' 'I'm really sorry [FORMAL]'.As a result, my friend texted 'You have to apologise so politely for such a trivial thing, [you make me feel like] [CASUAL]ne' no hitokoto de owaru yōna dekigoto deshita.Shikashi watashi wa, okorasete-shimatta to omoi, LINE de 'Saki ni kaette, kaban o hōttara kashi ni shite-shimai sumimasen deshita [FORMAL]' to teineina shazaibun o okurimashita.Tomodachi wa, 'Tomodachi ni tai suru taido ja nai' to henshin ga kimashita.Watashi wa sono kotoba no imi ga wakarazu, sono ato mo 'Sumimasen deshita [FORMAL]' 'Hontōni gomennasai [FORMAL]'