Conceptualizing Illocutions in Context: A Variationist Perspective on the Meta-Illocutionary Lexicon

The present study takes a metapragmatics approach to studying language users' conceptualizations of requests and apologies. In particular, two spoken sub-corpora within the BNC and the COCA were searched for inflectional and derivational variants of request and apology, which are part of the meta-illocutionary lexicon. The retrieved items were analyzed with respect to frequency of occurrence and communicative functions in context. Distributions of frequencies and functions were used as indicators of conceptualization, which were then compared across the two illocutions and across the two sub-corpora. The results of this study suggest that while cross-varietal differences are comparatively small, differences across the two illocutions are much more distinct. These and other findings are discussed against the background of semantic and pragmatic characteristics of these illocutions as well as relevant cross-cultural and social factors involved in discussing them. Finally, this study is characterized by its attempt to combine metapragmatics with variational pragmatics.


Introduction
Since Austin's (1962) introduction of illocutionary acts, pragmatics research has largely focused on formal realization patterns of speech acts or responses to them, either within a language (e.g. Manes & Wolfson, 1981), across languages (e.g. Chen, 1993), across varieties of a language (e.g. Barron, 2008), or across social factors such as socio-economic status (e.g. Rüegg, 2014). Together with the advent of metapragmatics, however, more recent studies have investigated how illocutions are being referred to by ordinary language users, which 1 3 communicative functions these references fulfill in discourse, and what the observations may reveal about the nature and laypeople's views on specific illocutions. A case in point is a study by Schneider (2017) introducing the concept of meta-illocutionary expressions (MIEs). MIEs are lexical items "used to talk about verbal communicative acts in spoken or written discourse, specifically to name, perform, negotiate, or discuss these acts" (Schneider, 2017: p. 229). Examples include threat, promise, or compliment, as well as all their inflectional and word class variants (e.g. threaten, promising, compliments, etc.). In his exploratory study, Schneider (2017), using a corpus of contemporary English prose fiction, established four potential communicative functions of MIEs, namely performative, reporting, commenting, and problematizing. However, there are no studies to the author's knowledge which examine quantitative distributions of communicative functions, which is especially regrettable given their potential value for conclusions about the nature and laypeople's views on illocutions.
The present study focuses on the illocutions of requesting and apologizing exclusively, and its overall aim is to provide an initial quantitative comparison of frequencies and communicative functions of the relevant MIEs. Since requests and apologies are essential for efficient communication in multiple languages and cultures (cf. Blum-Kulka, 1989b), examining both how and how often these illocutions are being talked about in discourse is argued to generate promising results regarding the apparent salience and transparency of the relevant illocutionary concepts, and the role these concepts play in (cross-)cultural and social understandings of verbal communication. To that end, this study adds a variationist perspective to the analyses, as results will further be contrasted across American English (AmE) and British English (BrE). MIEs were searched for in spoken sub-corpora of the Corpus of Contemporary American English (COCA) (Davies, 2008-) and the British National Corpus (BNC) (Davies, 2004-) respectively. Altogether, the present study was set up to answer the following research questions: 1. How often do MIEs characterizing requests and apologies occur in spoken English discourse? 2. Which communicative functions do they fulfill? 3. How do frequencies and functions vary across the requests and apologies? 4. How do frequencies and functions vary across AmE and BrE?
With respect to structure, the present paper will first provide an overview of previous research on requests and apologies while putting a special focus on realization patterns as well as their potential relevance to ordinary language users' conceptualizations of these speech acts. Subsequently, it will address the problematic framework of metapragmatics, review previous studies within this framework, and advocate a combination of metapragmatics and variational pragmatics. In the methodology section, sampling procedures concerning the two sub-corpora and the present coding scheme will be described in detail. After that, this paper will present and discuss frequencies of MIEs and, in a second step, turn toward their communicative functions. Results will be compared across illocutions have confirmed that the IFID sorry is "the overwhelming favorite" (Meier, 1998: p. 216), ranging from 79% of all apologies in New Zealand English (cf. Holmes, 1990) to 84% in spoken British English (cf. Aijmer, 1996). Together with other IFIDs such as pardon, excuse, afraid, apologise, forgive, regret, and apology, sorry belongs to a set of very few "routine expressions that are prototypically associated with the speech act of apology" (Lutzky & Kehoe, 2017: p. 40), which makes identifying them in discourse relatively easy (cf. Deutschmann, 2003: p. 36). The formulaic and routinized nature of apologies in use may reinforce the theoretical considerations presented above: due to the concurrence of clarity and politeness in terms of Lakoff's (1973) rules of pragmatic competence, a routinized but explicit apology may strengthen its face-saving effect and therefore generally appear more appropriate in context.
In contrast to apologies, the CCSARP distinguishes nine realization strategy types for requests (cf. Blum-Kulka et al., 1989b). These strategy types are grouped together as direct, conventionally indirect, and non-conventionally indirect (cf. Blum-Kulka & Olshtain, 1984: p. 201). Direct strategies include the mood derivable (e.g. Do X!) and performatives (e.g. I am asking you to do X.), conventionally indirect strategies include the query preparatory (e.g. Can/Could you do X?) or the suggestory formula (e.g. What about doing X?), and non-conventionally indirect strategies include hints (e.g. I am very cold as a request to close the window). The coding manual of the CCSARP has been applied in a wide variety of studies on requests, which found that this speech act is realized most often by conventionally indirect strategies, implying a balance between clarity and politeness (cf. Lakoff, 1973). This has been confirmed for British English (cf. Barron, 2008;House, 1989), Irish English (cf. Barron, 2008), and Australian English (cf. Blum-Kulka et al., 1989a, b), but not entirely for informal spoken American English, where requests tend to be more direct (cf. Flöck, 2016). Although the salience of conventionally indirect strategies seems to vary across many contextual factors, research has shown that the query preparatory may in some cases account for around 90% of all requests employed (cf. Barron, 2008: p. 55). All in all, then, a quite interesting picture emerges for requests and apologies: while apologies tend to be realized by routinized explicit expressions, preferably by the IFID sorry, requests are generally realized by conventionally indirect strategy types. Requests are thus not only realized more indirectly, but also by a wider variety of individual strategies within a strategy type or even a group of strategy types. These individual strategies may then vary in speakeror hearer-orientation, the use of modals, or different strategies of internal and external modification (cf. Blum-Kulka et al., 1989b).
The studies in pragmatics reviewed so far provide valuable insight into the theoretical nature of requests and apologies as well as their formal realization patterns. However, examinations of the ways these (and other) illocutions are being talked about in discourse are very scarce (cf. Schneider, 2017). Nonetheless, such examinations seem promising, as they may allow for conclusions about laypeople's implicit perceptions of the illocutions' natures as well as their salience and entrenchment in language users' minds. In the following section, two theoretical frameworks within pragmatics are combined, serving the 1 3 Conceptualizing Illocutions in Context: A Variationist… purposes of enabling both a meta-perspective on the study of illocutions as well as cross-varietal comparisons.

Metapragmatics and Variationist Perspectives
Although metapragmatics as a research area is quite heterogeneous, the main bulk of literature defines it as the "study of reflexive awareness on the part of participants in interactions, and observers of interactions, about the language that is being used in those interactions" (Haugh, 2018: p. 619). Apart from explicit and implicit instances of metacommunication about the present context, the study of reflexive awareness can also take as evidence ordinary language user's abstractions from interacting and assessments of pragmatic aspects of communication in general (cf. Hübler, 2011). Within the latter strand of research, we can distinguish the empirical-conceptual approach in metapragmatics, which examines, among other things, ordinary language user's understandings of concepts which, for the linguist, fall under the umbrella term of pragmatics. Broadly speaking, the aim of the empirical-conceptual approach is to arrive at conclusions about first-order understandings of second-order pragmatic concepts, and its methods mainly include lexical approaches and ordinary language users' talk about these concepts (cf. Caffi, 1998).
A conceptual domain within pragmatics which has attracted a considerable amount of attention in this context is the domain of speech acts, and particularly speech act verbs. For instance, Verschueren (1985) provides a comprehensive overview of speech act verbs dealing with the overarching topics of lying, silence, directing, and so-called forgotten routines. Taking a more comprehensive approach, Ballmer and Brennenstuhl's (1981) classification of speech acts comprises a detailed analysis of over 4,800 speech act verbs, and in a similar vein, Wierzbicka (1987) defines a somewhat smaller collection of such verbs using so-called semantic primitives as part of an allegedly universal Natural Semantic Metalanguage (NSM). What these studies roughly share is their methodological focus. Taking a primarily lexical approach to analyzing speech act verbs, the two studies often take fabricated contexts of use as evidence to substantiate their claims about the overall meaning of these verbs.
However, there has been a recent shift towards naturally-occurring data as the main source of evidence in analyzing metapragmatic labels. While Culpeper et al. (2017) focus on labels denoting concepts within (im)politeness, Schneider (2017) -most relevantly for the present purposes-conducts an exploratory study about the use of meta-illocutionary expressions (MIEs) in English prose fiction. MIEs are nouns, verbs, adjectives, and adverbs which denote some linguistic action (e.g. promise, congratulate, suggestive, apologetically) and are used by speakers to talk about an illocution on the meta-level as opposed to performing an illocution on the object-level. In this study, Schneider (2017) tentatively distinguishes four communicative functions of these MIEs, which denote how the respective illocutions are being talked about, evaluated, or assessed in context. The communicative functions in question were labeled performing/performative, reporting, commenting, and problematizing (cf. Schneider, 2017: p. 230).
In short, when used performatively, the MIE naming the speech act "is employed to actually perform this speech act" (Schneider, 2017: p. 230), which simultaneously characterizes the interface between the talk about a speech act (meta-level) and the performance of a speech act (object-level). When it is used in a reporting function, the MIE informs the recipient "that a particular speech act was performed in the more or less distant past or what the illocutionary force of an utterance was in the understanding of the speaker" (Schneider, 2017: p. 232). The commenting function is characterized by an MIE occurring in a speaker's retrospective comment toward their own previous utterance, in which the speaker clarifies the illocutionary force of, and their intention behind, this utterance (cf. Schneider, 2017: pp. 233-234). Finally, when used in the problematizing function, an MIE may either challenge the interlocutor's illocution itself or the legitimacy of this illocution due to, say, a specific social relationship between the interlocutors (cf. Schneider, 2017: p. 239). This function further encompasses requests for a speech act to be performed (such as in Promise?) and requests for confirmation that an illocution has been correctly identified (cf. Schneider, 2017: p. 235). While the categorization of communicative functions seems plausible at least at first sight, the author of that paper was not interested in examining quantitatively the distribution of these functions in discourse, which is precisely how the present study will complement the former.
In addition to examining communicative functions of MIEs in spoken English discourse, another aim of the present study is to explore variation across AmE and BrE. In that regard, this study is also situated in the framework of variational pragmatics. Variational pragmatics can be conceptualized as a field of research at the "interface of pragmatics with variational linguistics, i.e. with modern dialectology" (Schneider & Barron, 2008: p. 1), one of its major goals being to examine pragmatic variation across geographical varieties of a language. Since its introduction, this field has attracted much attention and continues to be researched to this day. However, hardly any studies seem to examine metapragmatic phenomena from a variationist perspective. Those studies that do take a variationist approach are mostly concerned with metapragmatic judgment data concerning directness and politeness degrees of speech act realization strategies (cf. Blum-Kulka, 1987;Fraser & Nolan, 1981). However, none of them seems to have focused on quantitatively investigating speaker's abstractions from interacting in discourse, let alone examining frequencies and communicative functions of MIEs. Being situated at the-as yet-largely overlooked interface of metapragmatics and variational pragmatics, this study seeks to fill this blatant research gap.

Data Collection
For the present purposes, the BNC and the COCA served as data collection instruments. More specifically, two sub-corpora were created within those corpora in order to achieve maximum validity. In the process, minor technical and practical limitations had to be taken into consideration as well. While it is argued that these limitations interfered only partly with the goal of attaining validity and comparability, they did certainly have an impact on the creation of the sub-corpora. The creation procedure is described in more detail below.
The first and certainly most obvious criterion for language data to be included into the two sub-corpora was that the data belonged to the spoken components of the corresponding corpora in the first place. Since both the BNC and the COCA were created with representativeness of the respective varieties of English in mind, it would seem reasonable at first glance to include the entire datasets from both spoken components into further analyses. However, one of the major differences between these two corpora-the diachronic nature of the COCA as opposed to the synchronic nature of the BNC-ties representativeness of these corpora down to two different temporal contexts. For the present purposes, it was argued that an accurate account of cross-varietal comparison presupposes a concordance of time frames from which the data originate. Therefore, a direct comparison of the spoken components of the COCA and the BNC would have invalidated the results due to temporal comparability issues. 1 In order to minimize said comparability issues while aiming for a large size of both sub-corpora at the same time, the spoken component of the BNC was adopted as a whole (first sub-corpus) and compared to selected data samples within the COCA collected between 1990 and 1992 (second sub-corpus).
While the 'sampling procedure' for the BNC sub-corpus does not require further attention, the corresponding procedure for the COCA sub-corpus resulted from a compromise between achieving maximum representativeness on the one hand and accepting minor technical limitations on the part of the online corpus software on the other. In particular, the initial idea of including all the spoken data between 1990 and 1992 into the COCA sub-corpus had to be dismissed due to an error message within the corpus software, which pertained to the size of this sub-corpus. On that account, it was decided that only the first 100 texts of the represented spoken subcategories (i.e. radio or television programs, e.g. ABC, CNN, PBS,...) for each year were going to be included into the sub-corpus. In those cases, the software automatically balanced the 100 texts across shows (e.g. ABC Primetime, ABC Jennings, ABC Brinkley,…). If, however, a sub-category was represented by less than 100 texts in a given year, all the available texts were included. Finally, every text which was part of this sub-corpus was taken over in its entirety from the COCA. Table 1 illustrates a more detailed composition of the sub-corpus in its final form.
In total, the BNC sub-corpus comprised 9,963,663 words compared to 6,635,541 words constituting the COCA sub-corpus. Needless to say, all absolute figures used to answer some of the research questions at hand were normalized and only then compared to each other and used as a basis for further inferences. Furthermore, the present study takes a synchronic approach to data analysis rather than a diachronic approach, mainly for the sake of having a larger set of temporally comparable data.
Taking a synchronic approach further partly compensates for the imperfect balance of texts across years and sub-categories within the COCA sub-corpus and maintains its representativeness of spoken AmE to a large extent. All in all, it is argued that the relatively high degree of representativeness of the two varieties-achieved through both sampling and a deliberate stabilization of the time factor-makes the two subcorpora eligible for exploring intra-lingual variation in the use of MIEs. However, it should also be noted that, despite some overlaps, the genres covered by the spoken components of the BNC differ from those of covered in the COCA. While the COCA covers news broadcasts exclusively, the BNC includes a wider variety of genres such as classroom talk, interviews, and public debates. Despite the exploratory nature of the present study, awareness should be raised to the fact that findings from the COCA sub-corpus may be more reflective of English 'broadcast talk' rather than spoken English in general.
In the present study, frequencies and communicative functions of MIEs were examined and compared across spoken AmE and BrE. The MIEs under consideration were word class and inflectional variants of apology and request. Since these two illocutions are very different in nature (see Sect. 2.1), it was hypothesized that a comparison of the corresponding MIEs would enable further insightful contrasts on a metapragmatic level. In particular, the nodes searched for in the subcorpora were apology, apologies, apologize/apologise, apologizing/apologising, apologized/apologised, apologizes/apologises, apologetic, and apologetically on the one hand, and request, requests, requesting, requested, requestive, and requestively on the other. Out of these nodes, apologetically, requestive, and requestively did not occur in any of the sub-corpora. The comparison of frequencies did not require any further data processing-the comparison of communicative functions was achieved through a coding procedure described in the following sub-section.

Coding
The scheme used to code communicative functions of MIEs was developed based on Schneider's (2017) categorization. While the performative, reporting, and problematizing functions were almost fully adopted for the present coding scheme, some notable modifications and additions were also made. These modifications and additions pertained to both labeling and contextual coverage of functional categories, which will be described in detail below. Despite being referred to as 'communicative functions of MIEs' throughout this paper, the pertinent categories actually denote the discursive functions of relevant utterances with respect to the MIEs they comprise. Colloquially speaking, these functional categories relate to what the utterances do to the MIEs. Altogether, six functions were distinguished: performative, problematizing, reporting, clarifying, naming, and commenting.
The performative function was almost fully adopted from Schneider (2017). The only extension that had to be made was that this function could also be realized by a group of people, or, grammatically speaking, using the first-person plural (e.g. We apologize).
As far as the problematizing function is concerned, this study also included all instances in which the problematization was aimed at an utterance performed by a person other than the speaker's immediate interlocutor(s). Example (1) is a case in point.
(1) In discussing the-the complaints that Iraq has with Kuwait, is-is he still apologizing for Saddam Hussein's actions.
The reporting function, otherwise also taken over entirely from Schneider (2017), was extended to instances in which the reported illocution occurred at the moment of reporting (realized through present simple or present progressive forms) or, counter-intuitively, even in the future. The reason for including references to prospective illocutions as instances of the reporting function was that the discussed utterances occurred in an institutional context. In example (2), the MIE occurred within an account of an ongoing official negotiation rather than within an instance of pure social or relational work between two speakers. As taken up below, similar instances of MIEs used in the context of relational work were coded as commenting.
(2) Tonight, Mr. Bush will request money for a variety of other existing health and welfare programs packaged under the umbrella of benefits for children.
The clarifying function only requires attention with regard to labeling, as this category is almost fully equatable with the commenting function introduced in Schneider (2017). The label clarifying was chosen for two reasons: first, it seemed more accurate than commenting, and second, commenting was used as a label referring to a different communicative function. Similar to the problematizing function, the clarifying function was extended to references to an illocution performed by a person other than the speaker, as shown in example (3). (

3) I think what he was doing, Sonya, was apologizing if it hurt somebody
personally.
The addition of the naming function arose from Schneider's (2017) statement that "MIEs are […] used to talk about verbal communicative acts in spoken or written discourse, specifically used to name [...] these acts" (2017: p. 229, emphasis added). Interestingly, Schneider (2017) did not introduce a corresponding category himself. However, naming illocutions seems to be common practice in spoken discourse (cf. Sect. 4.2) and is probably the most basic communicative function MIEs may fulfill. In particular, the naming function in this study encompasses all instances where the particular illocution appears as an external, representational denotatum which is not discussed, assessed, or negotiated any further. In a way, speakers simply echo other's statements and perceptions of illocutionary acts without necessarily endorsing them, often for reasons of communicative efficiency and intelligibility. Example (4) illustrates this category.
(4) Authorities in the United Arab Emirates refused the ship's request to enter the harbor at Dubai.
Finally, the commenting function comprised all instances which in some way implied a statement about how the corresponding illocution appears in context, how it is (not) used, should (not) be used, or will (not) be used. In short, this function broadly relates to interlocutors' perceptions of appropriateness as well as the illocutions' overall nature. Instances of this category included deliberate withholdings of the illocution (cf. (5)), references to a future performance of the illocution, often connected to a precondition (cf. (6)), explicit references to appropriateness (cf. (7)), references to a type of manner connected to the illocution (cf. (8)), or prompts to perform the illocution due to contextual factors (cf. (9)).
(5) I have nothing to apologize for.
(7) Thornburgh certainly should apologize for that and I-I believe he has. (8) What most parole boards might see as pandering, Doris Tate hopes is apologetic and sincere. (9) Give your apologies as you walk in.
A final important distinction to make, albeit on a higher level of abstractness, is between genuine metapragmatic uses and pseudo-metapragmatic uses of MIEs. While functional categories such as commenting, clarifying, or problematizing have to do with speakers' assessments and negotiations regarding a certain illocution, functions such as naming or reporting are characterized by merely mentioning the illocution in question and thus simply echoing others' functional perception of an utterance. Consequently, the former categories are more reflective of the on-line salience of respective illocutions in language users' minds, whereas 1 3 Conceptualizing Illocutions in Context: A Variationist… the latter-or a combination of the two-are more strongly related to cognitive entrenchment and the idea of prepackaged linguistic concepts (cf. Schmid, 2010).
As is the case with many coding schemes, there were a couple of instances in the data which could have been coded more than one way. Especially the clarifying, the problematizing, and, to a degree, the commenting function seem somewhat interrelated. For reasons of presentiveness and practicality, relevant examples were coded as instances of the categories which, according to the author's own estimation, were semantically and pragmatically closest to the respective utterance. Finally, a few instances which could not be categorized due to insufficiency of contextual information were coded unclear.

Frequencies of MIE Use
Frequencies of MIE use are argued to largely reflect the degree of entrenchment of the corresponding illocutions in ordinary language users' minds. More precisely, these frequencies illustrate how often language users use prepackaged linguistic concepts to make themselves and their interlocutors aware of the illocution of a particular utterance. Tables 2 and 3 contrast absolute and normalized frequencies of the respective illocutions' nodes along the lines of the two varieties. A quite clear pattern can already be spotted at first sight, both across illocutions and across varieties. The following presentation and discussion of results will at first provide a cross-varietal comparison of total frequencies across illocutions and, in a second step, zoom in successively on each of the two illocutions and compare frequencies of individual nodes.

Comparing Illocutions
By comparing normalized total frequencies of MIEs across illocutions and varieties, a surprisingly distinct trend can be identified. All frequencies (except one) are quite consistent across illocutions and varieties, showing that MIEs account for roughly 30-35 items per million words in the respective sub-corpora. Against this background, MIEs characterizing requests in AmE occur almost twice as often (χ 2 ≈ 39.09; p < .00001 for absolute frequencies), which points especially toward crossvarietal differences in the entrenchment of request as a concept 1 in AmE and BrE speakers' minds. Such differences do not exist in the context of apologies, as frequencies of MIEs are strikingly similar. This particular distribution of frequencies allows for multiple conclusions to be made and diverse potential explanations to be given. First, and most obviously, requests as a concept seem to be more entrenched in AmE than in BrE speakers' minds. Second, among AmE speakers, the concept of  requests also seems to be more entrenched than the concept of apologies. This does, however, not hold true for BrE speakers, among whom these two concepts seem to be almost equally present. It is argued that this particular distribution of frequencies, and therefore the different degrees of entrenchment of specific illocutions, may be influenced by crossvarietal differences in discursive practices involving metapragmatic terms. In particular, it may be the case that AmE and BrE speakers have a different understanding of the semantic field of request, which is reflected in the use of MIEs in discourse. In fact, the concept of requests lends itself quite easily to differing interpretations because the boundary of what counts as a request, especially as opposed to other directives such as orders or commands, is rather blurry. One may, for instance, understand order as a specific type of request which would prototypically be very direct and face-threatening. However, one may also intentionally separate these two concepts and distinguish them along the lines of specific speaker-hearer relationships: for example, one could argue that an order may be issued only by superiors in an institutional context such as by one's own boss at work or by municipal authorities, and that directives of a similar kind in non-institutional contexts would merely count as (direct) requests. The fuzzy boundary of requests is also reflected in the finding that request generates the most Google hits in the context of selected structural variants indicative of an MIEs' problematizing function (cf. Schneider, 2017: p. 237). Following this argumentation, then, the presented frequencies may reveal that AmE speakers use request as a cover term for a range of different directives, which, in turn, would explain the more frequent activation of the mental concept of request in AmE. Conversely, BrE speakers may classify the mental concept of request into the same level of concreteness as other directives, which may lead to a more precise gradation of directives in BrE speakers' minds, and, ultimately, to comparatively lower frequencies of MIEs characterizing requests in BrE. However, this last claim clearly requires further evidence.
The lack of cross-varietal differences in the context of apologies is in accordance with both the above argumentation frame and findings in Schneider (2017). In contrast to directives, which all imply a certain degree of verbally getting the hearer to do something (cf. Searle, 1976: p. 11), expressives convey feelings which are prototypically easy to distinguish based on the respective illocutions performed by a speaker. Therefore, while using request as a hyperonym for request, order, and command may indeed seem plausible, using apology as a cover term for, say, apology, thanks, and congratulations would not make much sense. Accordingly, the MIE apology generates the least Google hits as part of specific structural variants indicating a problematizing function of an MIE (cf. Schneider, 2017: p. 237). Since the semantic coverage of apology is not as open to interpretation as is the coverage of request, cross-varietal differences in frequency may be less likely to occur, which is reflected in corresponding frequency values.

Comparing Nodes: Apologies
Comparing frequencies of individual nodes within a specific illocution such as apology sheds light on the use of MIEs' forms in interaction. However, since there are hardly any one-to-one relationships between forms and functions of MIEs (cf. Sect. 4.2), mere comparisons of the nodes' quantitative distributions do not seem exquisitely revealing and will thus be kept rather short here. Without a doubt, these distributions do indeed become highly relevant if mapped onto communicative functions of MIEs, but unfortunately, systematic analyses of this kind lay beyond the scope of the present study. Nonetheless, initial observations regarding potential tendencies in form-function relationships of MIEs will be briefly mentioned in the attempt to explain the distribution of frequencies across individual nodes.
Frequencies of some individual nodes vary significantly, even though accumulated frequencies of MIEs characterizing apologies are very similar across AmE and BrE. Most prominently, when normalized, APOLOGIES occurred roughly four times as often in BrE than in AmE (χ 2 ≈ 24.33; p < .0001 for absolute frequencies). Conversely, APOLOGY occurred about twice (χ 2 ≈ 10.37; p ≈ .0013), and APOLOGIZED more than three times as often in AmE than in BrE (χ 2 ≈ 22.32; p < .0001). Some variation could also be found for APOLOGIZE, which was more frequent in BrE, but this variation is of minor magnitude compared to the aforementioned nodes and not statistically significant (χ 2 ≈ 0.728; p = .3935). The remaining nodes occurred rather infrequently and showed no substantial variation across the two sub-corpora.
Based on preliminary observations regarding potential form-function relationships, the remarkable salience of APOLOGIES in BrE seems to be due to crossvarietal differences in performing apologies. While Apologies, My apologies, or Our apologies as actual realizations of apologies were found quite often in the BNC subcorpus, they were rather rare in the COCA sub-corpus. In addition, BrE speakers also performed apologies by saying I apologize and We apologize more often than AmE speakers, which could, amongst other things, explain the higher frequency of APOLOGIZE in BrE. Based on these findings, one could hypothesize that AmE speakers correspondingly prefer Sorry, I am sorry, or We are sorry (cf. Flöck, 2016;Lutzky & Kehoe, 2017), but this claim requires verification.
By contrast, the relative predominance of APOLOGY and APOLOGIZED in AmE may be intuitively explained by a potential preference for using MIEs characterizing apologies in the naming and the reporting function respectively. Indeed, this seems to be the case for APOLOGIZED and the reporting function, which, verifying previous observations of this kind (cf. Schneider, 2017: pp. 232-233), do seem to correlate in the present dataset. Regarding APOLOGY and the naming function, the picture appears much less clear. In fact, APOLOGY fulfilled many different communicative functions in both sub-corpora, meaning that this node is seemingly less prototypical of the naming function than initially expected. Consequently, this finding also precludes further plausible explanations for the relative predominance of APOLOGY in AmE, which could be a potential aim for following studies.

Comparing Nodes: requests
While the accumulated frequencies of MIEs characterizing requests vary substantially across the two varieties, relative frequencies of individual nodes are distributed almost equally. In fact, all normalized values for both individual and accumulated 1 3 frequencies of nodes display a ratio of roughly 2:1 in favor of AmE. This finding is in stark contrast to the context of apologies, where accumulated frequencies were almost the same, but frequencies of individual nodes showed notable variation. The only values that slightly deviate from the aforementioned ratio are those pertaining to REQUESTS and REQUESTED, but, as implied, these deviations are marginal and not statistically significant (χ 2 ≈ 0.552, p = .4575 and χ 2 ≈ 0.458, p = .4985 respectively). Rather than cross-varietal differences, these minor deviations may just as well reflect variation across individual instances of data collection (i.e. shows, meetings, speeches). What is more, the node REQUESTS is particularly ambiguous, as it may either represent a third person singular verb in the present tense or a plural noun. REQUESTED, just as APOLOGIZED, is also ambiguous, as it may represent the past tense of a verb or the past participle. Therefore, specific contrastive conclusions about the use of these two nodes must be withheld at this point.
As initial observations and forthcoming results suggest (cf. Sect. 4.2.2), individual nodes of MIEs characterizing requests are much more indicative of their communicative function than those characterizing apologies, and this holds true for both varieties. With regard to REQUEST and REQUESTS, the latter statement applies along the lines of word class. REQUEST and REQUESTS as nouns mostly fulfilled a naming function, whereas REQUEST and REQUESTS as a verb, REQUESTING, and REQUESTED tended to fulfill a reporting function. Deviations from this pattern did, of course, occur, but they were surprisingly rare. Taking into account that relative frequencies across nodes are strikingly similar and that individual nodes seem to correlate highly with communicative functions at first sight, it appears all the more interesting that the accumulated frequencies of nodes vary so substantially across varieties. In fact, these findings combined reinforce the claim that request as a mental concept is much more entrenched in AmE speakers' than in BrE speakers' minds, possibly because the semantic coverage of this concept is so open to interpretation. However, although this explanation of results seems to be very promising in the present context, it needs to be treated with caution. Differences in genre, which do exist across the two sub-corpora, may have also had an influence on cross-varietal use of MIEs, but again, a more detailed account on this issue requires more research.

Communicative Functions of MIEs
Quantitative distributions of MIEs' communicative functions are argued to reflect how ordinary language users perceive the corresponding illocutions' natures and, more precisely, their roles and functions in interaction. The following presentation and discussion of results will at first address the context of apologies and then turn to the context of requests. Within both contexts, relative frequencies of communicative functions of MIEs will be contrasted across varieties and explained by ordinary language users' subconscious understandings of the particular illocutions. Figure 1 illustrates relative frequencies of communicative functions of MIEs which characterize apologies. Needless to say, reference values for the calculation of percentages are the accumulated frequencies of nodes per variety as illustrated in Table 2, meaning that each node fulfilled exactly one communicative function. As can be seen in Fig. 1, nearly all communicative functions display a notable degree of variation across AmE and BrE. Most prominently, MIEs in performative function occurred almost twice as often in the BNC sub-corpus than in the COCA subcorpus. 2 With regard to the commenting function, the proportions are inversed, the function being found almost twice as often in AmE than in BrE. The reporting function accounted for 28.5% in the AmE and 22.7% in the BrE data, 3 which, taking into account the overall salience of this function across both datasets, represents a comparatively small degree of variation compared to other functions. The clarifying and the problematizing functions occurred quite infrequently in the two datasets, but both functions were still notably more salient in AmE. The naming function occurred almost twice as often in BrE as in AmE but occurred comparatively infrequently across the two sub-corpora. Finally, there were a few instances in BrE in which the MIEs' functions could not be identified.

MIEs Characterizing Apologies
What appears quite striking in this distribution is that only two of the functions established in Schneider (2017), namely performative and reporting, occur frequently in spoken English discourse. By contrast, clarifying (based on commenting from Schneider, 2017, cf. Sect. 3.2) and problematizing functions occur conspicuously rarely in both present datasets, which raises the question of general applicability of Schneider's (2017) categorization. It has to be borne in mind in this context that Schneider (2017) worked with prose fiction and therefore used a philological method of data collection, which "can only cover a limited amount of material and therefore necessarily must remain very selective" (Jucker, 2009(Jucker, : p. 1616). Moreover, since prose is a specific kind of art, it prototypically involves what can be colloquially called an idealized and cleansed use of language, which is why "[f]indings for fictional language obviously cannot be generalised to other forms of language" (Jucker, 2009(Jucker, : p. 1616. Clearly, while the categorization in Schneider (2017) may be of avail for investigating MIE use in fictional data, it can merely serve as a point of reference for spoken data.
Apart from the sparsity of clarifying and problematizing functions, it appears to be a general trend that MIEs which characterize apologies most often perform, comment on, report an occurrence of, and occasionally also name this illocution. However, as outlined above, cross-varietal differences across functions are substantial. The difference in the use of the performative function appears to be closely connected to the divergent salience of APOLOGIZE and APOLOGIES across BrE and AmE. Example (10) is just one of very many instances in which APOLOGIZE correlated with the performative function in BrE. Of course, no claims can be made about the general frequency of performing apologies across varieties, mainly because speakers of AmE may, for example, simply prefer other realizations of apologies such as the IFID sorry as suggested by the initial strategy in example (11). 4 What can, however, be concluded is that speakers of BrE, as opposed to speakers of AmE, clearly prefer MIEs in realizing apologies, which itself may be related to differing politeness norms in the respective cultures. Another reason for this difference between BrE and AmE may be the choice of corpora. It is conceivable that apologizing is less common in broadcast talk than in other genres of spoken discourse, which is why the performative function may display a higher frequency in the BNC sub-corpus compared to the COCA sub-corpus.
(10) But as I said, I do apologize.

(11) And what I've done was wrong and I lied and I'm sorry for it, and I apologize to everybody
As mentioned above, the commenting function is much more salient in the AmE data than in the BrE data. The commenting function comprises all instances where an MIE was used to assess the appropriateness of, the need for, or the nature of the respective illocution, and it can already be concluded from Fig. 1 that AmE speakers perform more utterances of this kind than BrE speakers. In other words, AmE speakers, more often than BrE speakers, tend to verbalize under which conditions apologies are appropriate and how their nature is expressed in discourse.
In contrast to the performative and commenting functions, which either actively put into effect or assess the contextual function of an apology, the naming and the reporting functions do not touch upon the immediate effect of an apology on interpersonal relationships. These functions rather objectify the illocution and treat it as a static entity: MIEs in these functions either simply assign an illocutionary label to some action or refer to the (unproblematic) existence of such an illocution within a past course of events. Across varieties, however, these functions were distributed differently in quantitative terms, which was mostly due to divergent compositions of the respective sub-corpora. In particular, AmE news reports caused comparatively more reporting functions, whereas in BrE, business meetings triggered comparatively more naming functions. Naming functions were prototypically realized through noun forms, and reporting functions through verb forms of MIEs, but there were many exceptions to this pattern. The exchange in example (12), taken from a British business meeting, shows that the node APOLOGIES often fulfilled both naming and reporting functions: this context shows one of many instances where the chairman of a meeting names the illocution and thereby prompts a report, and the person in charge complies by reporting apologies for absence from other virtual members of the meeting. Taken together, it appears quite striking that functions such as performative or commenting, which specifically relate to the effect of apologies on interpersonal relationships and the overall communicative context, are found more frequently than functions such as naming, which objectify this illocution and make it appear unproblematic. Two aspects are argued to be responsible for this distribution, one relating to apologies as an isolated concept and the other one to perceived contextual preconditions of apologies. First, apologies, by their illocutionary nature, always convey some sort of hearer-centered feelings in order to restore a social relationship (cf. Searle, 1976: p. 12), which would explain the salience of the performative function. The explanation for the salience of the commenting function is more complex: while an apology itself may be easily identified as an apology by ordinary language users, notions of contextual preconditions for an apology may differ substantially, which is why interlocutors may use the commenting function to negotiate their views on the appropriateness of, or the need for, an apology. In particular, apologies require a past act initiated by the speaker and perceived hurtful for the hearer by either the speaker, the hearer, or both, and each of these aspects may be subject to negotiation. Taking into account percentaged distributions across AmE and BrE, it seems that altogether, AmE speakers put a stronger focus on negotiating preconditions of apologies, whereas BrE speakers tend to emphasize the illocution's courteous nature in discourse.

MIEs characterizing requests
With respect to MIEs which characterize requests, Fig. 2 illustrates the distributions of their communicative functions across AmE and BrE. Without a doubt, distributions depicted in Fig. 2 represent a stark contrast to those shown in Fig. 1, both in terms of overall salience of functions and in terms of cross-varietal differences. With respect to overall salience, results show a clear, and almost one-sided, tendency towards the use of MIEs in either the reporting or the naming function. In fact, the two functions combined account for roughly 90% of all functions in both varieties. Performative and commenting functions make up most of the remaining 10%, and individual occurrences of clarifying, problematizing, and unidentifiable functions complete the picture. What is more, there is hardly any variation across AmE and BrE: the only tendency is that BrE speakers use performative and commenting functions slightly more frequently, whereas AmE speakers more often tend to report and name the illocution. Otherwise, the distribution is remarkably similar across varieties, which may hint at a certain degree of universality in English speakers' perceptions of requests. As discussed in detail below, it may be the case that requests are perceived static and not as sensitive to contextual preconditions as apologies, which again relates to the two illocutions' theoretical natures.
As far as formal realizations of communicative functions are concerned, the pattern is largely similar to the one observed above (cf. Sect. Comparing nodes -apologies). The performative function was most often realized through REQUEST as a verb, the reporting function through REQUESTED as simple past or past participle form, or REQUESTING as a continuous form, the naming function through REQUEST or REQUESTS as nouns, and the commenting function through all of the above. Of course, exceptions to this pattern were found, but they seemed to be even rarer than in the context of apologies. Thus, it really appears to be the perceived illocutionary nature of requests, rather than differences in formal MIE use, that account for this distribution of communicative functions.
Since requests are preferably reported or named in both AmE and BrE, they seem to be perceived as static entities whose interactional goals appear unproblematic for speakers. This claim is reinforced by the fact that MIEs in the problematizing function occurred even less frequently for requests than for apologies, which, however, directly contradicts the findings in Schneider (cf. 2017: p. 237). Of course, Schneider's (2017) findings are based only on specific formal variants of the problematizing function rather than the whole function itself. Nonetheless, the enormous quantitative discrepancy between request and apology found in his study was argued to reflect that apologies have a routinized realization pattern and are therefore easily identifiable in discourse, whereas requests are often realized more indirectly and consequently less identifiable. Thus, taking into account the present study's contradictory finding concerning the problematizing function on the one hand and the overwhelming salience of the naming and the reporting functions on the other, an alternative explanation is offered in the following, both emphasizing the nature of illocutions and addressing the use of MIEs in institutional as opposed to non-institutional discourse.
With respect to illocutionary nature, emphasis will be put again on the role of prototypical contextual preconditions of performing an illocution. More precisely, especially in contrast to apologies, requests do not require as specific a context in order to be performed. For performing requests, it is sufficient for the speaker to merely believe that a hearer can perform the requested act and will probably only do so if actually requested, whereas for performing apologies, a past act is required which has to be perceived harmful for the hearer by either the speaker, the hearer, or both. In more abstract terms, requests only presuppose one party's estimation of the other party's potential ability and behavior, whereas apologies require a specific constellation of both party's actual or hypothetical attitudes toward a past event. Accordingly, the fact that there are many more contextual preconditions for apologies to be performed means that potentially more factors may be open to negotiation by interlocutors. For example, speakers may disagree on the person responsible for, the perceived harm caused by, the intention behind, or even the actual existence of a past event, and each of these topics may be subject to value judgments expressed through the MIEs' commenting functions. By implication, then, the reason for requests being less susceptible to MIEs' comments about appropriateness and practice of use seems to be the smaller number of contextual preconditions tied to the eligibility-or felicity, for that matter-of this illocution. That is not to say, of course, that speakers need not attempt to achieve appropriateness in requests by choosing a certain degree of indirectness according to macro-social and micro-social factors, for instance. Rather, the results are argued to reflect that due to comparatively few contextual preconditions, requests tend to be identified and talked about more effortlessly in discourse.
Another possible influence on the present distribution of communicative functions may be the peculiarity of the sorts of discourse in which labels such as request are used. However, it has to be mentioned in advance that this argument is fully based on intuition and therefore requires empirical evidence. As it seems, MIEs characterizing requests tend to occur often in institutional discourse, but not as often in non-institutional discourse such as small talk or conversations among friends.
In these types of non-institutional discourse, constructions such as X asked me to do Y or instances of broadly paraphrasing request with question may easily replace the use of MIEs. Of course, question is not an illocution, but it can be argued that ordinary language users do not distinguish between the two labels and therefore use them interchangeably. In institutional discourse, however, request seems to be used in any context where some good or action is being asked for. It may thus be the case that request is perceived to be part of erudite language by ordinary language users and therefore used more often in institutional contexts, which in turn made up most of the data in the present sub-corpora. Again, even though these arguments may indeed apply, specific empirical support is urgently needed.

Conclusion
The present study investigated frequencies and communicative functions of MIEs in spoken English discourse. The focus lay on MIEs characterizing requests and apologies, and results were compared across AmE and BrE. MIEs were searched for in spoken sub-corpora of the COCA and the BNC respectively, and manual micro-analyses were employed to determine their communicative functions in context. With respect to frequencies, results indicate a clear entrenchment of MIEs characterizing requests in AmE, which occurred almost twice as often as all remaining groups of MIEs, both across illocutions and across varieties. This finding was ascribed to the unclear semantic coverage of request, which, according to ordinary language users' interpretations, may or may not encompass other illocutions such as order or command. It thus may be the case that AmE speakers conceptualize request differently than BrE speakers by using it as a cover term for a multitude of directive acts. With regard to frequencies of individual nodes, it was found that APOLOGIES and APOLOGIZE occurred more often in BrE, whereas APOLOGY and APOLOGIZED were preferred by AmE speakers. Preferences in BrE were argued to reflect British politeness norms in performing an apology, and preferences in AmE were expected to hint at AmE speakers' preference for using MIEs in naming and reporting functions respectively. Notably, in the context of requests, no cross-varietal preferences for individual nodes could be found. In line with accumulated frequencies, each individual node roughly displayed a frequency ratio of 2:1 in favor of AmE.
Distributions of communicative functions depict distinctive variation across illocutions as well as notable cross-varietal differences within individual communicative functions. First of all, MIEs which characterize requests are overwhelmingly used in the naming and reporting function across both varieties, whereas those characterizing apologies more often perform and comment on this illocution. AmE speakers make more use of the commenting function, whereas BrE speakers prefer the performing function. Reporting and naming functions are also used in the context of apologies, but they occur more rarely, especially in relation to the performing and the commenting functions. This striking variation across illocutions is argued to be related to their respective contextual preconditions. In particular, apologies require that a very specific event had happened in the past, that this event was in some way harmful for the hearer, and that this event was actually perceived harmful for the hearer by one or both conversational partners, making each of these aspects open to value judgments and negotiation. Requests, on the other hand, only presuppose the speaker's belief concerning the hearer's ability and future behavior, which is why they may be identified and referred to more unproblematically in discourse. Furthermore, it was suggested that the use of request may prototypically be bound to institutional discourse in which illocutions are not normally commented on or problematized, but rather seen as a means to advance work-related issues and therefore just named or reported.
In face of all its telling results and valuable implications for research on the nature of illocutions, this study also struggled with shortcomings, which were mostly of methodological nature. First of all, even though the COCA and the BNC both claim representativeness of the corresponding varieties, they are composed differently from one another. While the COCA sub-corpus almost exclusively comprises radio or TV news broadcastings, the BNC sub-corpus also includes a large number of recorded meetings, which may have distorted the comparability of the two sub-corpora to a certain degree. Secondly, the choice of the time period, which was a necessary compromise made due to different temporal settings covered by the COCA and the BNC, is far from ideal because the analyzed data was collected almost 30 years ago, and needless to say, language-and also MIE-use may have changed since. Finally, it may have been fruitful to further subdivide the commenting function in the coding scheme because even though all instances did, in fact, comment on the nature, the practice of use, or the appropriateness of illocutions, they did so in quite different ways. A more detailed coding scheme could thus enable more precise conclusions regarding MIE use as well as language users' perspectives on the nature of illocutions. Guiding ideas of the present study clearly need to be complemented or expanded upon.
Most obviously, upcoming studies could, for example, look at other illocutions and scrutinize the present claim regarding the importance of illocutions' contextual preconditions. Moreover, as called for above, the coding scheme could be modified or improved in order to make more detailed and valid claims about quantitative results. Other than that, further research could focus on written instead of spoken language, on other varieties of English, or even on MIE use across languages. Anyhow, combining the frameworks of metapragmatics and variational pragmatics seems very promising, as this may further disclose how interactional functions of specific linguistic actions are perceived cross-culturally by ordinary language users.
Funding Open Access funding enabled and organized by Projekt DEAL. No funding was received for conducting this study.

Conflict of interests
The author has to relevant financial or non-financial interests to disclose.

Human and animal rights statment
The research reported on in the following did not involve human participants or animals.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.