Referring to Speech Acts in Communication: Exploring Meta-Illocutionary Expressions in ICE-Ireland

This paper presents an exploratory study of meta-illocutionary expressions, i.e. of everyday vocabulary such as request, promise and insult, used to explicitly refer to speech acts and to label, discuss and negotiate them in communication. Taking the example of expressions referring to acts of apologising in Irish English, the question is addressed in which contexts, forms and functions these meta-illocutionary expressions are used. Evidence from ICE-Ireland and SPICE-Ireland is employed to demonstrate how this question can be answered empirically, thus providing a model for a corpus-based analysis of such metapragmatic folk terms and how they are used in spoken and written discourse, across genres and illocutionary types. This paper thus offers a blueprint for investigating the meta-illocutionary lexicon of ordinary language users and their first-order conceptualizations of communicative action.


Introduction
This paper presents a study in metapragmatics which is focused on the meta-illocutionary lexicon, i.e. on lexical items such as request, promise and insult, used to refer to speech acts and available to all competent speakers of English. These metaillocutionary expressions (henceforth MIEs) are part of the everyday vocabulary and commonly used to talk and write about linguistic action. These words are employed by ordinary language users, who are not at all familiar with speech act theory and do not know such technical terms as 'illocution' or 'metapragmatics'. Meta-illocutionary expressions are not technical terms, but folk terms, i.e. labels employed by 1 3 lay people to categorize utterances in terms of what language philosophers and linguists call 'illocutionary acts', and the underlying category concepts are (to employ a distinction first adopted in politeness studies, cf. e.g. esp. Eelen, 2001) first-order conceptualisations, i.e. lay persons' interpretations and understanding, as opposed to second-order conceptualisations, i.e. theoretical constructs and technical definitions provided by experts in language philosophy and linguistics.
In the present study, the overarching question is: When, how and why are metaillocutionary expressions used? This question concerns the following aspects: (1) When are speech acts made the topic of discourse? How frequent are references to a particular speech act? Are MIEs used in spoken or written language? Which discourse genres are they used in? These questions also entail matters of frequency and distribution, e.g. whether MIEs are used more often in speech or in writing, and also whether some speech acts are referred to and explicitly discussed and negotiated more often than others. Findings regarding this last issue may have implications concerning the relative salience of speech acts in the minds of language users and their importance in interpersonal relations.
(2) How are MIEs employed in discourse? Which manifestations are used, in terms of word class (e.g. verb or noun) and word form (e.g. past or present tense)? What types of constructions are they part of? Are there typical collocations? A further relevant parameter may be discourse position, i.e. where in discourse MIEs typically occur (e.g. at the beginning or the end of a conversation or a letter).
(3) Why are MIEs used in communication? What are their functions? Which speech acts do they occur in, and which speech acts do they refer to (e.g. in They promised to come, the verb promise is a MIE referring to a promise, but occurring in an assertion)? And what are the underlying conceptualisations?
With this focus on the meta-illocutionary lexicon and the uses of meta-illocutionary expressions, the present study is not interested in the related fields of speech act recognition and speech act realization. There is work on speech act recognition and speech act processing in conversation, predominantly by psychologists (e.g. Holtgraves, 2008;Licea-Haquet et al., 2019), and there is much more work on speech act realization in applied linguistics (e.g. Blum-Kulka et al., 1989;Trosborg, 1995). Yet these areas are outside the scope of the present paper.
For the purposes of demonstrating the approach introduced here, the examples chosen are MIEs which can be used to refer to apologies, e.g. the noun apology or the verb apologise. This paper is not, however, about apologies as such, nor about the different ways available to realise and perform apologies in English (although performatives will play a role at some stage, cf. Sections 2.2 and 4.5.2), nor scholarly definitions of apologies or studies of this speech act. In more general terms, this paper is not about speech acts, but about speech act labels. Or, to put it even more bluntly, it is not about speech acts, but about words.
In the following section, the present study is situated in the field of metapragmatics (Section 2.1), before the meta-illocutionary lexicon is introduced and discussed (Sect. 2.2). Section 3 presents a five-step procedure for a corpus-based analysis which may serve as a blueprint for investigating the meta-illocutionary lexicon. Some findings resulting from the deployment of this procedure are offered in Sect. 4. In particular, details are given about frequencies (4.1), the distribution of word classes and word forms (4.2), and the distribution across speech and writing (4.3), across text categories (4.4), and across speech act classes (4.5). The last subsection (4.5.4) provides a glimpse at conclusions which can be drawn from the study of MIEs about a lay person's understanding of a given speech act, in this case apologies, underscoring the potential legal relevance of such conclusions. The paper closes with a brief summary in Sect. 5.
A definition of metapragmatics relevant to the present purposes is included in Culpeper and Haugh (2014: 239), who write that metapragmatics "concerns the use of language on the part of ordinary users or observers, which reflects awareness on their part about the various ways in which we can use language to interact and communicate with others". It is important to note that the focus is on lay people (i.e. on their first-order concepts) and not on language philosophers or linguists (i.e. on their second-order concepts). As Jakobson (1960: 356) observes when introducing the metalingual function as one of the six basic functions of language: "[…] metalanguage is not only a necessary scientific tool utilized by logicians and linguists; it plays also an important role in our everyday language." In this spirit, the present paper concentrates on one particular aspect of everyday language reflecting ordinary language users' pragmatic awareness, and specifically their awareness of speech acts as the basic units in communication (Searle, 1969: 16). Yet the focus is not on how speech acts can be realised and produced, but on how they are labelled and discussed. It is crucial to notice this difference. So, the focus of the present paper is, in other words, on the vocabulary used to explicitly refer to speech acts, i.e. on the meta-illocutionary lexicon, and not on the speech acts themselves, in this case not on acts of apologizing, but first and foremost on nouns and verbs used in everyday 1 3 communication to refer to acts of apologizing. Apologies have been chosen for demonstrative purposes alone to illustrate the type of investigation advocated here.
Since the speech act labels focused on in the present study have been classified as 'folk metalanguage' (Culpeper & Haugh, 2014: 239) and the underlying first-order concepts as 'folk categories' (Verschueren, 1985) (cf. Section 2.2 below), folk pragmatics would also seem a relevant field. Yet, Niedzielski and Preston (2009), for example, explicitly distinguish their notion of folk pragmatics from metapragmatics, differentiating between, on the one hand, referring to particular utterances by using labels such as polite, which for them is metapragmatics (and which is the focus of the present study), and, on the other hand, discussing more extensively language attitudes and beliefs e.g. about "good language", which for them is folk pragmatics (which is not what the present study is focused on).

The Meta-Illocutionary Lexicon
The term 'meta-illocutionary lexicon' collectively refers to all words in a language that can be used to explicitly name speech acts, or more precisely that most crucial component part of speech acts termed 'illocutionary acts' by both Austin (1962) and Searle (1969). Examples of meta-illocutionary expressions (MIEs) include request, promise and insult. MIEs may be verbs or nouns (the three examples presented here are ambiguous in this regard).
In speech act theory, there has been an obsession with verbs. In Austin's approach, so-called performative verbs play a pivotal role, specifically 'explicit performative verbs' (e.g. Austin, 1962: 61). These are verbs which can be used to name and at the same time perform the speech act they name. For instance, in 'You will see it before anyone else, I promise.' (McEwan, 2013: 310;cf. the literary sources at the end of the reference section) the verb promise both names and performs an act of promising. Yet not all meta-illocutionary verbs can be used performatively. Threaten and insult, for example, do not work in that way. They can only be used to name, but not to perform the respective speech act. Mey (2001: 105-111) distinguishes performative verbs from speech act verbs. Other authors also prefer this broader term. Wierzbicka, for instance, published a dictionary of English speech act verbs (1987). While she includes approximately 250 common verbs, Ballmer and Brennenstuhl (1981), aiming at a comprehensive lexical analysis, examine 4,800 verbs, which they refer to as 'English speech activity verbs'. Verschueren (1985) compares English and Dutch speech act verbs, but also at least mentions speech act nouns.
Experiments on speech act recognition show that language users are able not only to recognise speech acts, but also to name these acts, and for this purpose they use nouns, not verbs. For example, most users say that Shall we go to the movies? is a suggestion, and I'm really sorry is an apology (Schneider, 2019). Such experiments show that MIEs belong to a lay person's 'folk metalanguage' (Culpeper & Haugh, 2014: 239) and that the underlying concepts are 'folk categories' (Verschueren, 1985), which may or may not coincide with the second-order conceptualisations developed by scholars and defined e.g. in terms of felicity conditions (Austin, 1962). Indeed, Edmondson (1981: 139) addresses this point when he notes:

3
Referring to Speech Acts in Communication: Exploring… we face a difficulty of nomenclature in attempting to catalogue illocutionary acts for our analytic purposes. Wilful avoidance of everyday terms seems unusually presumptuous, but to take over commonly-used terms such as 'promise' may be totally misleading.
Apart from meta-illocutionary verbs and nouns, there are also meta-illocutionary adjectives and adverbs, as the following examples demonstrate, which are taken from contemporary English novels by employing the 'philological method' (Jucker, 2009(Jucker, : 1616, i.e. extracted manually from fictional material. (1) He apologised for taking his time. (McEwan, 2013: 200) (2) […] Justin recognised this statement to be a sort of apology, […] (Aw, 2013: 89) (3) He had caught the hostility in my voice and his touch on my arm was tentative, or apologetic. (McEwan, 2013: 72) (4) "Hmm, I steamed the broccoli, but it is still rather firm," her mother mumbles apologetically. (Guo, 2015: 306) Based on such evidence, it has been argued that MIEs appear in a range of different communicative functions, each typically, though not exclusively, correlating with a specific constructional pattern (Schneider, 2017). The functions tentatively suggested are: (a) a performative function (cf. performative verbs mentioned above), typically expressed by a present simple form of meta-illocutionary verb with first person singular reference (example 5); (b) a reporting function, typically expressed by a past simple form with third person singular reference (example 1 above); (c) a commenting (or clarifying) function, typically expressed by a progressive form with first person singular reference (example 6); (d) a problematizing (or challenging) function, typically expressed by a Yes/No question with a meta-illocutionary noun (example 7) (for details, cf. Schneider, 2017: 230-236).
(5) "[…] I do apologise, this must be very tedious for you." (Kennedy, 2016: 321) (6) "[…] I wasn't apologising, I was saying sorry because I didn't hear you." (Kennedy, 2016: 109) (7) "Is that a threat, Mal?" (Black, 2006: 131, orig. emphasis) Even though no example of the problematizing function referring to an apology was found by employing the philological method, the typical construction, in four variants (Is/Was that/this an apology?), is very frequent on the internet. A quick Google search carried out in 2016 yielded approximately 1 million hits, and with other MIEs up to 11 million hits (Schneider, 2017: 237). More generally, some MIEs are among the most frequent English words. The ICE-CORE lemma list, for instance, which is based on the vocabulary used in seven different ICE corpora, includes the following eight MIEs (with decreasing frequencies): order, offer, suggest, request, invite, promise, warn and threat. These eight thus belong to the approximately one thousand most frequent words of the English language, accounting for 70-80 % of the words used by fluent speakers of English (for further information about ICE-CORE, cf. e.g. Gilner & Morales, 2011). Therefore, these eight MIEs can be considered most prominent among core speech act concepts (on this notion, cf. Verschueren, 1985;Wierzbicka, 1987).

Method
All examples discussed in section 2.2 were gathered manually from prose fiction. While this philological method doubtlessly has its merits, like any other method, and is well suited for qualitative study to generate hypotheses, it cannot be employed for systematic quantitative analysis (Schneider, 2018: 72-73). Hence, to answer the questions posed in the introduction, a corpus-based approach is adopted here instead. The corpus chosen for this purpose is ICE-Ireland (Kallen & Kirk, 2008), i.e. the Irish component of the International Corpus of English (ICE). This corpus, like all ICE corpora, includes approximately one million words, of which 40% are written and 60% are spoken language. Additionally, this corpus displays a unique feature not found in other ICE corpora. Of the spoken data, there is a pragmatically annotated version, SPICE-Ireland (Kallen & Kirk, 2012), permitting e.g. an analysis across Searle's five speech act classes (Searle, 1976).
For the corpus-based analysis of MIEs in ICE-Ireland, a five-step procedure is suggested. For all conceivable word forms of all relevant word classes of MIEs used to explicitly refer to a particular speech act, the following characteristics have to be established: (i) Overall and relative frequencies; (ii) Distribution of word classes and word forms; (iii) Distribution across speech and writing; (iv) Distribution across text categories; (v) Distribution across speech act classes.
The results permit inferences regarding the underlying conceptualisations. To illustrate the five analytical steps, the example of references to apologies is chosen. While apology or apologise are not among the eight MIEs included in ICE-CORE (cf. section 2.2 above), acts of apologising have received a tremendous amount of attention in the research literature (for an overview, cf. e.g. Oishi, 2013;also Rieger, 2017: 558-562), probably as much as requests, and definitely considerably more than any of the speech acts labelled by the remaining seven MIEs in ICE-CORE.

Frequencies and Salience
MIEs for referring to apologies can be verbs, nouns, adjectives or adverbs, as illustrated in examples 1-4 above. These MIEs can occur in twelve different word forms, spelling variants included. These are: apologise, apologize, apologises, apologizes, apologised, apologized, apologising, apologizing, apology, apologies, apologetic and apologetically. These forms are collectively represented as APOLOGY. Employing the search form apolog, a total of 46 instances was found in ICE-Ireland. This may seem to be a low frequency, yet references to requests are not much more frequent, with a total of 52 instances (request occupies rank 4 among the eight MIEs in ICE-CORE, cf. Section 2.2 above). In fact, half of these eight MIEs occur in ICE-Ireland with a total of fewer than 100 instances in each case. The largest number of occurrences is found for OFFER, a total of 291 (offer occupies rank 2 among the MIEs in ICE-CORE). In other words, acts of apologising are explicitly referred to in Irish English discourse only slightly less frequently than other core speech acts. By contrast, references to some of the 250 core speech act concepts included in Wierzbicka (1987) occur just once (e.g. assert, bemoan, profess) or not at all (e.g. bewail, confide, exult). Clearly, some speech acts are talked and written about more often, and thus seem to be more salient, than other speech acts.

Distribution of Word Classes and Word Forms
Of the 46 APOLOGY instances, 28 were forms of the verb, and 18 were forms of the noun. There were no adjectives or adverbs. All verb forms, without exception, were spelled with -ise; spelling variants with -ize did not occur (it has, however, to be borne in mind that spelling of the spoken transcriptions in ICE-Ireland was homogenised into a standardised British orthography; cf. Kallen & Kirk, 2008: §9). Three different forms of the verb were found, namely 18 occurrences of apologise, 8 occurrences of apologised, and 2 occurrences of apologising. Among the 18 nominal forms were 13 occurrences of apology and 5 occurrences of apologies. This means that only five of the twelve possible word forms occur in ICE-Ireland (or five of the eight possible forms, not counting the spelling variants). Apart from adjectival and adverbial forms, the verb form apologises (3 rd ps. pres. simple) is not attested.

Distribution Across Speech and Writing
The next step in the analysis of MIEs concerns their distribution across speech and writing. 31 instances of all word forms of APOLOGY occur in the spoken data, and 15 in the written data (24 when normalized per 600,000 words, i.e. the approximate size of all spoken sub-corpora, for immediate comparison across ICE corpora). This distribution is relatively balanced, with the majority of occurrences in spoken discourse (56.4%). This is quite different from the distribution of the eight MIEs included ICE-CORE (cf. Section 2.2 above). For each of these, without exception, a clear majority of all instances in ICE-Ireland occur in written discourse, on average approximately two thirds, while only one third occurs in spoken discourse (Schneider, 2021).
The distribution of the individual APOLOGY word forms is summarized in Table 1. The numbers from the smaller written sub-corpus are normalized and rounded for direct comparison. The raw numbers appear in parentheses, the totals in bold type.

3
The numbers show that apologies are more often explicitly referred to in spoken language than in written language. Only the plural form of the noun occurs more frequently in the written data. Table 1 also shows that the verb forms outnumber the nominal forms in the spoken sub-corpus (21 versus 10), while the distribution is more balanced in the written sub-corpus. In both speech and writing, the dominant verb form is apologise, and the dominant noun form is apology.

Distribution Across Text Categories
In step iv of the analysis (cf. Section 3), the distribution of the MIEs across discourse genres is examined. In ICE-Ireland, a total of 32 "text categories" is distinguished, 15 for spoken and 17 for written language. The former include e.g. Telephone conversation, Legal cross-examination and Scripted speeches; the latter e.g. Business letters, Creative writing and Press editorials (cf. Kallen & Kirk, 2008: 9). Table 2 shows the distribution of the APOLOGY MIEs across these categories.
The 46 MIEs occur in fewer than half of all text categories, namely in 14 out of all 32. The 31 spoken MIEs are used in 8 of the 15 spoken categories, and the 15 written MIEs in 6 of the 17 written categories. Most spoken MIEs are used in an institutional context: 13 in legal discourse (Legal presentations: 9, Legal crossexamination: (4)), 8 in broadcasting genres (Broadcast discussion: 5, Broadcast talks: (3)), 3 in Unscripted speeches, delivered by a Minister of State (2) and a chief executive, and 1 in a Parliamentary debate. Only 6 MIEs occur in conversation (Face to face: 5, Telephone: (1)). In the light of these findings, it seems that explicit reference to the act of apologising (and, possibly, explicit reference to speech acts more generally) is more necessary and hence more frequent in institutional discourse than in private contexts.
The 15 written MIEs are used predominantly in letter writing (8 instances). 3 occur in Social letters, and 5 in Business letters, i.e. in an official context, supporting the assumption that explicit mentioning and negotiation is more typical of official than of private contexts. 3 further instances occur in Press news, 2 in Creative writing, and 1 each in the categories Popular technology and Skills and hobbies. Table 2 also shows that the corpus employed is relatively small and, hence, the number of instances of the MIEs is low. Yet, some interesting observations can be made. First, the verb forms occur overwhelmingly in spoken text categories. This applies to apologise in particular (13 versus 5 instances). Second, in 5 of the 8 spoken categories, but only 1 of the 6 written categories, no noun forms are used. On the other hand, there are 2 written text categories, but no spoken categories, in which no verb forms are used. Third, in Face to face conversation, apologised is used more frequently than in any other text category, and also more frequently than any other word form in conversation. This seems to suggest that in conversation apologies are typically reported (cf. Section 4.5.1). This may also be true of the 2 occurrences of apologised in the category of Press news. Fourth, the most frequent form apologise is spread comparatively evenly across the text categories. It occurs in 10 of all 14 categories, specifically in 7 of the 8 spoken and 3 of the 6 written genres. Given that, at least in the spoken data, MIEs are used predominantly in official and institutional contexts, this may mean that apologise 1 3 is typically used not only to refer to the act of apologising but to actually perform acts of apologising (cf. Section 4.5.2).

Distribution Across Speech Act Classes
In step v. of the analysis (cf. Section 3), the distribution of MIEs across speech act classes is examined. The focus here is on the occurrence of the expressions in the spoken data in the five illocutionary types annotated in SPICE-Ireland, i.e. in Representatives (REP), Directives (DIR), Commissives (COM), Expressives (EXP) and Declarations (DECL). Table 3 shows the distribution of the verb and noun forms across these five speech act classes. The 31 spoken word forms of APOLOGY occur in only three of the five speech act classes. They are most frequent in Representatives (17 instances), and are also used in Expressives (8) and Directives (6), but not in Commissives or Declarations. In Representatives, all five word forms -the three verb forms and the two noun forms -occur. In Expressives, on the other hand, only three forms are used (apologise, apologising and apologies), whereas in Directives only two are used (apologise and apology). Put another way, only apologise occurs in all three speech act classes (REP, EXP and DIR), apologised in only one (REP), and the three remaining forms in two classes, namely both apologising and apologies in REP and EXP, and apology in REP and DIR.

APOLOGY in Representatives
The spoken form apologised is used exclusively in Representatives, in which this form occurs six times altogether. In all cases, it is employed to report that an act of apologising had occurred, i.e. used in the reporting function (cf. section 2.2 above). As a rule, the apologiser is a third party, prototypically a male individual (in five of the six cases), alternatively the speaker themselves. The following is a typical example: (8) <rep> So he called and he apologised and all </rep> <P1A-044$A> All six apologies are reported as part of a narrative, relating past events. This applies to the four instances from face-to-face conversation (including the above example) as well as the two remaining instances respectively occurring in a broadcast talk and in a legal cross-examination. This furthermore applies to the two instances of apologised in written language (not included in SPICE-Ireland). Both these instances occur in a newspaper article and appear at the very beginning of this article, specifically in the lead following the headline (example 9), and in the closing sentence of this same article (example 10).
(9) <p> <#> <bold> A Ballymote man who claimed he was <quote> "hassled" </quote> by a Garda had summonses against him dismissed when he apologised before Boyle District Court and when the court heard that he had a hearing problem. </bold> </p> <W2C-016$C> (10) <#> However, the summonses were later dismissed when Leonard apologised to the court, and when it was later discovered that the defendant had a hearing problem. </p> <W2C-016$C> Interestingly, the apology reported in this newspaper article was made in a legal context, as is the case for the apology referred to in the cross-examination, in example 11.
(11) <,> he presented later on a letter of apology in which he apologised to the government <,> <P1B-069$B> This example, in which apologised co-occurs with letter of apology, is a reminder that apologies do not only occur in spoken discourse, as much of the research in pragmatics seems to suggest, but also in written discourse, and can be performed in an official letter of apology, as is the case in the political context reported in this legal cross-examination. This example underscores the fact that explicit reference to apologies is characteristic of institutional discourse (cf. Table 2 above), and that apologies can be legally relevant. This is even true for the apology reported in the broadcast talk, which refers to the killing of a man <P2B-027$A>. Irrespective of the context, the basic construction in all cases is he apologised, merely stating that an apology was made (as in example 8). Variants include the addressee of the reported apology (e.g. he apologised to the government) or the offence (e.g. apologised for killing an innocent man). If information about addressee or offence is not included in the utterance containing the MIE, then it is known from the broader context this utterance is produced in.
The noun form apology occurs in Representatives as frequently as the verb form apologised, but five of the six instances of apology appear in the very same situation, in the same file, together with the two instances of apology in Directive acts. In short, seven of all eight occurrences of apology in the spoken data are used in the same speech event, which is a legal presentation (cf. Section 4.5.4 for details). The eighth instance is the letter of apology in example 11 above. This means that all instances of apology in the spoken data are found in legal discourse.
Of the five instances of apology in written discourse (and hence not included in SPICE-Ireland), two occur in Expressives (cf. Section 4.5.2.), the other three occur in Representatives. One of these three is used in a social letter to report an apology of the letter writer (<W1B-010$A>). In the other two cases, the apology is not only reported but also evaluated as insincere and unsuccessful. Example 12 is taken from a creative writing text, and example 13 from a press news text.
(12) <#> The soldier bent double and squeezed some feeble words of apology between the whinnies of his laughter. <#> <W2F-002$A> (13) <p> <#> The request was declined and the Parry's refused to meet privately because they were not prepared to have a <quote> "mumbled apology from the likes of Gerry Adams over the mistakes of Warrington". </quote> </p> <W2C-002$D> In example 12, the third party apology is all the less convincing as it co-occurs with laughter and is thus an insufficient remedy of the (immaterial) damage done by the apologising soldier. The same qualifier feeble also combines with metaillocutionary nouns designating other speech acts, as in the following example from the Belfast Telegraph. The source issuing the excuse referred to is the Shell company. Their excuse is categorized as very feeble.
(14) <p> <#> "To say people here wouldn't be interested in half-price ferry travel is a very feeble excuse. <#> <W2C-006$E> In example 13 (which also includes the meta-illocutionary noun request), the apology referred to is only expected to occur, and expected to be so insincere that the offended party are not prepared to actually grant the potential apologiser an opportunity for apologising. This reported event from a political context reveals an attitude towards public apologies which seems to be wide-spread (cf. e.g. Battistella, 2014). There are many discussions of apologies by politicians, e.g. on social media and in discussion forums, in which such apologies are perceived, or at least strategically assessed, as insufficient and therefore qualified in a number of different ways, including the use of mumbled, or referred to as nonpologies, and hence rejected.
Here are examples posted on Twitter in response to what Taoiseach Enda Kenny said initially about the involvement of the Irish state in the Magdalene Laundries (institutions for "fallen women" run by the Catholic church and supported by the state) (https:// www. thejo urnal. ie/ twitt er-react ion-magda lene-laund ry-report-783949-Feb20 13/).
(17) An expression of regret and sorrow is no apology.
These tweets were posted on 5 February 2013. Two weeks later, a formal state apology was issued. The Taoiseach performed this apology by using the meta-illocutionary verb, intensified with unreservedly and combined with a verbal expression of regret, intensified with deeply (Dáil Oireachtas Éireann, 2013-02-19): (18) Therefore, I, as Taoiseach, on behalf of the State, the government and our citizens deeply regret and apologise unreservedly to all those women for the hurt that was done to them, and for any stigma they suffered, as a result of the time they spent in a Magdalene Laundry In this particular example, the meta-illocutionary verb apologise is not used to report or evaluate an apology, but to explicitly perform it. So in this case, the metaillocutionary expression does not occur in a Representative, but in an Expressive.

Apology in Expressives
In the Expressives identified in SPICE-Ireland, there is a total of eight APOLOGY occurrences, six of the verb form apologise, and one occurrence each of apologising and apologies. All six occurrences of apologise are used as IFIDs, deployed in the performative function for explicitly realizing acts of apologising. In each case, the grammatical subject is the first person singular pronoun, as in example 19 from a legal presentation.
(19) <exp> I apologise for that My Lord@ </exp> <P2A-064$A> While performative realizations prototypically involve the classical performative formula I apologise, the plural form of the meta-illocutionary noun can also be used in this function, resulting, however, in a much less formal apology, as in example 20 from a legal cross-examination. In this particular case, the apologiser is the judge, whereas in example 19, the apologiser is a barrister. The hierarchical difference may account for the different performatives chosen and thus the level of formality.
(20) <#> <exp> 1ApOlogies% </exp> <P1B-061$A> (<#> Apologies <S1B-061$A>) The only instance of apologising in an Expressive co-occurs with an instance of apologise in the same unscripted speech. Both instances are used in the introductory part of this speech. After a number of Representatives and Expressives, specifically acts of thanking, the speaker, a Minister of State, produces the utterance in example 21, which, arguably, can be interpreted as a hedged performative realization of an apology, yet in that particular context it functions rather as an announcement of an apology (and could therefore, alternatively, be annotated as a Commissive act). This utterance initiates a discourse unit which is closed by example 22, the apology proper. Between these two acts, the speaker gives the reasons for the delay he apologises for.
(21) <exp> I should also start by apologising for being late </exp> <P2A-025$A> (22) <exp> So I 'm uh <,> I apologise for that delay </exp> <P2A-025$A> In addition to the occurrences of apologise in the spoken sub-corpus, there are five instances of apologise in the written sub-corpus, which are also all used in the performative function. One is found in a popular technology journal, in what looks like a footnote: "<*> asterisk </*> The Editor wishes to apologise sincerely to Mr Connell for allowing […]" <W2B-040$A>. In this case, the apology is intensified with sincerely, a slightly weaker version of unreservedly used in example 18 (cf. also example 27). The formal character of this published apology is increased by using a hedge ("wishes to apologise") and more particularly the fact that the apologiser deploys third person self-reference. A second instance is found in a short story, represented as direct speech in the context of an everyday conversation: "'I do apologise Mrs Kilroy… […]'". Here an alternative intensifier is used. It is used because it took the speaker some time to recognize his interlocutor. The remaining three instances of apologise in written Expressives occur in business letters, where they are used to perform an apology in the final paragraph, before the formal sign-off. These apologies, which can be referred to as 'sealing apologies' (cf. House, 1982), are realised using formulaic language, with only minimal variation.
(23) <p> <#> I apologise for any inconvenience this may cause. </p> <W1B-022$E> (24) <p> <#> I apologise for any inconvenience that this decision may cause. <#> <W1B-016$E> (25) <p> <#> May I apologise for any inconvenience you have been caused. </p> <W1B-018$H> The written sub-corpus also includes five instances of apology, of which three occur in Representatives (cf. Section 5.4.1.), the remaining two in Expressives. These two Expressives are included in letters, of which one is a social letter and the other a business letter. In the social letter, apology is used without modification (example 26), while in the business letter it is used in a conventional formal phrase with the intensifier unreserved (example 27; cf. also example 18).
(26) <#> I will begin with an apology merited by the fact that I have totally failed to keep in contact. <#> <W1B-007$A> (27) <p> <#> Needless to say you should NOT have been mailed a "Business" promotion; and for that you have my unreserved apology. </p> <W1B-019$B> Both these examples illustrate that using the meta-illocutionary noun in its singular form, apology, is a further option to explicitly perform acts of apologising, resulting however in more elaborate versions than is the case with the performative formula involving apologise, and especially with the elliptical plural form of the noun, apologies. Needless to say, studying MIEs in a performative function overlaps with the study of speech act realisation, as MIEs in this particular function are used to name and to perform a speech act at the same time.

3
Referring to Speech Acts in Communication: Exploring…

Apology in Directives
Finally, four instances of apologise and two instances of apology are the only word forms found in Directives. Two instances of apologise occur in overlapping turns in a broadcast discussion in which a politician (B) is advising a company director to apologise, while the broadcaster (A) is asking whether the director is willing to apologise (example 28). A similar request is included in example 29 from a broadcast talk. The fourth instance of apologise also occurs in a request for information, in the context of a face to face conversation between friends. Speaker B asks whether her interlocutor, speaker A, remembers somebody else trying to apologise.
(30) <dir> Cos do you remember at the Christmas party last year <,> John tried to apologise to him and he brought him a drink Ronan said+ <quote> thanks </quote> and just put it on the table and walked away </dir> <P1A-070$B> Apparently John's attempt at apologising was not successful, and speaker A does not think that it even counted as such, as she says in her response: (31) <rep> Oh* John didn't try and apologise </rep> <#> <rep> He tried to <,> make amends </rep> <P1A-070$A> It is not reported what John had actually said nor why speakers A and B think that what he said was at best an attempt at apologising, or not even that. It would be interesting to know which criteria have to be met for them to accept utterances as apologies proper. Such criteria emerge in a slander case included in ICE-Ireland (4.5.4.).

What is a Fulsome Apology?
The study of MIEs not only provides insights on the frequencies and distribution of word forms across text categories and speech act classes, but may also throw light on the underlying conceptualisations of the acts referred to by these expressions, in the present case concepts of apologies, as will be briefly demonstrated by examining an excerpt from a slander case. The file "Slander case" <S2A-061> includes a legal presentation involving two barristers (A and C) and a judge (B). The plaintiff is a 46-year old district nurse, wrongly accused of not paying for petrol. The defendants are the manageress of the petrol station, as the main accuser, and her assistant. The plaintiff's case is that some customers in the station overheard the accusation, which damaged her reputation. The following discourse unit is initiated by the judge, asking for confirmation that an apology had been made. Barrister C's answer begins with a Representative which includes four further occurrences of the seven instances of apology in the entire dialogue.
(32) <S2A-061$B> <#> There was an apology <#> Is that right <S2A-061$C> <#> Well it it is <,> it <,> the defendants drafted it in terms that there was an apology uhm the plaintiff 's case uhm is that an apology in rather grudging terms was made not by the manageress of the shop who was the main accuser <,> but in fact by the assistant who had come over to her <#> But she would say that the apology was made at the end of this uh rather lengthy and unpleasant conversation uhm and it was one which was made grudgingly and it was not a fulsome apology as one would have expected <#> The barrister's answer culminates in saying that while an apology was made, it was not made properly, i.e. "it was not a fulsome apology". It is not known what was actually said, but at least three features of the apology were not "as one would have expected": (a) the apology was made too late, after a "rather lengthy and unpleasant conversation"; (b) it was made by the wrong person, i.e. not by the main accuser but her assistant; and (c) it was not made sincerely and convincingly, but "in rather grudging terms", "grudgingly". Because of these shortcomings, the apology was not accepted. It has to be borne in mind, however, that rejecting the apology in this court case can be a purely strategic move, worth a considerable amount of money. Results such as these add to the host of findings about apologies in general and the (in)sincerity of apologies in particular (e.g. Haugh & Chang, 2019).
The "Slander case" example as well as other occurrences of meta-illocutionary APOLOGY expressions mentioned above highlight the fact that apologies and how they are conceptualized can be legally relevant and may require explicit discussion or negotiation. The same is true for other speech acts, most prominently promises, insults and threats (e.g. Muschalik, 2018;Schneider & Zielasko, 2017).

Conclusion
The present paper includes an exploratory study of meta-illocutionary expressions (MIEs), i.e. explicit references to speech acts, in discourse. Following an earlier qualitative manual analysis of written fictional material, a quantitative corpus-based analysis of naturally occurring spoken and written data is outlined. For this study, ICE-Ireland was chosen, using references to apologies as an example. The analysis outlined focuses on five parameters: (i) overall and relative frequencies; (ii) distribution of word classes and word forms; (iii) distribution across speech and writing; (iv) distribution across text categories; (v) distribution across speech act classes.
Major results can be summarized as follows: (a) A total of 46 MIEs referring to apologies were found, 28 verb forms and 18 noun forms. (b) 3 different forms of the verb were used (18 occurrences of apologise, 8 of apologised, and 2 of apologising), and 2 forms of the noun (13 occurrences of apology, and 5 of apologies). (c) 56.4% of all instances occurred in spoken discourse, and 43.6% in written discourse, while for references to other core speech acts the average ratio is approximately 30 versus 70%. (d) MIEs were used in 14 of the 32 text categories distinguished in the corpus, mostly in institutional contexts (especially legal discourse and broadcasting). (e) Most MIEs are used in Representatives (in a reporting function); they are also used in Expressives (in a performative function) and Directives (e.g. in requests for information or clarification).
These results must not be overrated, as the frequencies of MIEs analysed here were relatively low, pointing to the comparatively small size of the corpus employed. Yet, while the quantifications presented have to be treated cautiously, a systematic analysis of naturally occurring speech and writing in a wide range of text categories has nonetheless been carried out, if only to illustrate the suggested approach. Finally, it should be emphasized here that the type of analysis demonstrated above opens a window on first-order conceptualisations of speech acts, as sketched out in Sect. 4.5.4. MIEs and how they are used in discourse for negotiating speaker intentions and hearer comprehension may help to identify properties of utterances, and the circumstances under which they are produced, revealing what lay persons consider necessary for an utterance to count as a 'proper' instantiation of a particular speech act, which may in some cases even be legally relevant, as illustrated.
Obvious options for future research are many. Meta-illocutionary expressions used to refer to further speech acts, e.g. requests, promises and threats, should be examined in ICE-Ireland for immediate comparison and also, like MIEs referring to apologies, in larger corpora. MIEs for any given speech act can be studied across varieties of English by using other ICE corpora, which are immediately comparable, thus contributing to variational metapragmatics (Haugh, 2018;Schneider, 2021). Furthermore, since one half of the data in ICE-Ireland was collected in Northern Ireland, and the other half in the Irish Republic, this corpus also permits an analysis of regional variation, thus making a further contribution to variational metapragmatics possible. And finally, MIEs can of course be investigated across languages (contrastive metapragmatics) and in the history of a language (historical and diachronic metapragmatics; cf. relevant publications by Jucker and Taavitsainen, e.g. 2014). In addition to these research perspectives, the use of MIEs in fictional material also seems to merit further study, as some authors use MIEs extensively, especially in prose fiction, while other authors do not use them at all. This applies in particular to deployment of MIEs as verbs of saying (verba dicendi), as e.g. in "This hurts," she complained. versus "This hurts," she said. For the purposes of studying such phenomena specialized machine-readable corpora are also available.