2.1 Theoretical reflections and methodological assumptions

Regarding the model of metadiscourse developed by Hyland / Tse (2004), Hyland (2005) and Hyland / Wang / Jiang (2022), the distinction between ‘interactive’ resources on the one hand and ‘interactional’ on the other⁶ implies an analysis on two levels: the level of discourse organisation (1), i.e. the formulation of “a connected discourse” (Hyland / Wang / Jiang 2022: 2), and the level of interaction (2), i.e. the expression of the writer’s (or speaker’s) “attitude towards either what or who is addressed” (ibid.). Examples for (1) are in other words; Furthermore, …; First …, then …; the second point is; in conclusion; to sum up. Examples for (2) are It is unfortunate that; I argue / think / believe … that; possible; might; perhaps; obviously; highly. The forms and specific discursive functions of these devices vary greatly with respect to the situation, genre-specific conventions, and overall discursive strategies. Metadiscursive resources can be linked to modality, e.g. in combination with modal verbs (e.g. should, could etc.), to topicalization (e.g. Whether this is true [is not certain]) or the use of questions to attract the reader’s/ hearer’s attention (e.g. Why is this a remarkable feature?). On the level of formulation strategies, the use of expressions of perception and cognition, such as hear, notice, see; feel; think; understand can be analysed in terms of metadiscursive strategy, in particular in evaluative contexts such as It is a comfort to hear … or It is important to understand … They represent ‘constructions’ on the basis of their relative fixity due to lexico-grammatical constraints exerted by the underlying respective patterns, thus providing immediate access to their meaning regarding their (meta-)discursive frame:

Construction (1) refers to a discursive frame in which the speaker reacts to a piece of information by expressing their satisfaction. The contrast between the formal, impersonal formulation It is … to hear of the Verb Phrase (VP) and a Noun Phrase (NP) expressing a strong positive emotion implies a certain degree of cognitive distance of the speaker with respect to the information. The interactional meaning will have to be negotiated with the interaction partner in order to determine the authenticity of the sentiment thus expressed. Regardless of the degree of irony, the stylistic effect plays out on a metadiscursive level in the sense that it “express[es] the writer’s attitude towards either what or who is addressed” (Hyland / Wang / Jiang 2022: 2, see above). The combination of the impersonal formulation It is … with the perception verb to hear functions as an additional device which invites the addressee to share their point of view. The interaction partner’s understanding of this meaning is based on a complex process of explicature and implicature (see e.g. Sperber / Wilson 1993: 12–14, 21–23), which enables them to react coherently by either confirming or rejecting the attitude thus formulated (Yes, indeed! / No, really?).

Construction (2) differs from the first example insofar as it contains both interactive and interactional resources: By pointing the addressee’s attention to a new aspect of the discourse, it acts as a connector providing textual coherence. Simultaneously, the evaluative expression important represents an ‘attitude marker’, following the definition by Hyland (2005: 180): “Attitude markers indicate the writer’s affective, rather than epistemic, attitude to propositions, conveying surprise, agreement, importance, frustration, and so on, rather than commitment” [Italics in the original].

The variety of descriptive and analytic categories employed in metadiscursive studies is considerable⁷. The categories regarding stance-taking, for instance, are frequently linked to specific grammatical categories and modalities, such as hedges, based on adverbs, adverbial or modal constructions (e.g. possible, might, perhaps). In addition, the difficulty in describing the complexity of text production and interactional strategies leads to an almost infinite number of potential linguistic candidates to be classified within the category of EoO. Hence, other theoretical aspects will have to be included in the definition of this notion with a view to giving a more precise account of the linguistic phenomena which I would like to take into consideration for the purposes of teaching and learning German as a Foreign or Second Language⁸.

The intimate link between genre and metadiscursive strategies have been repeatedly pointed out by various researchers (e.g. Firdaus / Soemantri / Yuliawati 2021; Hyland / Wang / Jiang 2022). This link plays an important role within a text analytical approach based on a definition of ‘style’ as “acting in a socially meaningful way” (“sozial relevante (bedeutsame) Art der Handlungsdurchführung”, Sandig 1986: 23). I thus propose to adopt a pragmatic approach to text and discourse in terms of ‘speech act patterns’, thereby following the German text analytical tradition of “Textmuster” (i.e. ‘textual patterns’). According to this tradition, genre can be specified with reference to different Textmuster as a structured unit based on a network of patterns regarding intonation, sentence and text structure, knowledge, style and Handlung, i.e. the different ways we (inter-)act linguistically (cf. Sandig 1989: 133). It is further assumed that expressing one’s opinion represents a speech act pattern which is connected to other speech act patterns within argumentative discourse, such as describing, illustrating, justifying, encouraging, expressing regret⁹. In this analytical framework, genre is intimately linked to the notion of ‘text type’, which means that it can be specified with respect to different textual patterns, such as reader’s comment, editorial, academic conference, press review.

As regards the speech act pattern ‘expressing one’s opinion’, it is prototypically connected to a series of text types. A list of useful expressions for learners of GFL provided by the Goethe Institute may serve as an example. The document can be downloaded directly from the internet and is conceived as a ‘learning aid’ (“Lerntipp”) enabling the learner “to write discussion essays, comments, or for an oral comment on a subject”¹⁰. The expressions listed in this document are referred to as “Redemittel”, i.e. ‘discursive devices’¹¹, and can serve as a perfect illustration of the metadiscursive distinction between organisational and interactional devices discussed above. However, the document does not merely present a list of linguistic constructions of varying complexity (e.g. Diese kurze Beschreibung führt zu der Frage …; Für … spricht …; Auch wenn …, meine ich doch, dass …). The learner is also given advice concerning the sequential organisation of their text/speech as well as the norms and conventions regarding argumentative discourse in general, for example: “Eine Stellungnahme muss/sollte eine klare Gliederung und Strukturierung in eine Einleitung, einen Hauptteil und einen Schluss haben”¹². The explicit references as to how argumentative discourse is expected to be organised, combined with equally explicit references to specific genres/text types (i.e. discussion essays, comments, oral comments) show a relatively high degree of awareness regarding both text types and text production strategies. Despite the awareness-raising qualities of this document, it is unsure whether the above-mentioned indications of text types enable the learner to differentiate sufficiently between these types and to use the different devices appropriately.

2.2 ‘Expressing one’s opinion’ in the light of ‘meaning’

The question as to which EoO may be appropriate regarding specific discursive genres/ text types requires a granular analysis of a series of different text types and text production strategies, which is beyond the scope of the present exploratory study. Nevertheless, it can be noted that the notion of appropriateness plays an essential role if we seek an answer to why French students of GFL frequently use expressions like *in meine Meinung or *meiner Meinung nach ist, dass … in their written productions. In both cases, we tend to qualify these errors as ‘mistakes’ by imputing them either to lexico-grammatical interferences with other foreign languages learned simultaneously – in this case, *in meine Meinung could be imputed to a false analogy with in my opinion in English – or else to insufficient morphosyntactic knowledge leading to ‘agrammatical’ constructions. From a construction analytical point of view, however, it can be argued that the distinction between ‘mistake’ and ‘inappropriateness’ is fallacious in the sense that they represent two sides of the same coin. By placing these errors on a scale within the grammar-lexicon-continuum, the analysis of errors in terms of either purely grammatical or lexical ‘mistakes’ is of secondary concern.

Within a constructionist framework, the meaning-aspect of language and language learning becomes a major pedagogical and didactic issue. A particularly salient example for the application of a meaning-orientated model of description is the usage-based, holistic Phraseoframe (Schafroth / Imperiale 2019), a canvas for lexicographic description of familiar idiomatic expressions in Italian, integrated into an online learner platform associated to the GEPHRI-project¹³. In combining frame semantic theory (cf. e.g. Fillmore 1982; Croft 2001; Ziem 2008; Busse 2012) with basic theoretical and analytical principles in Construction Grammar, this approach aims at enabling non-experts to understand phrasemes in terms of constructions, i.e. complex symbolic units or linguistic signs whose meaning emerges from the ways in which they are used in various contexts¹⁴. The Phraseoframe does not merely gloss the meaning(s) of a particular phraseme in combination with corpus-based examples, it also provides information and guidelines concerning their functions and use in specific contexts, including pragmatic and situational aspects¹⁵. Even though this dictionary does not contain any metalinguistic discursive devices comparable to the ones discussed in the present study, quite a few phrasemes are directly or indirectly linked to a semantic frame which could be referred to in terms of an ‘opinion frame’. For example, the meaning of the phraseme non stare né in cielo né in terra, (word-for-word-translation: *‘be neither in heaven nor on earth’) is glossed as follows: “(oft über Äußerungen, Gerüchte, Ideen o.Ä.) [überaus] absurd, unsinnig oder unzutreffend sein [und somit haltlos oder untragbar werden]”¹⁶. The suggested attribution to the ‘opinion frame’ appears justified on at least two accounts:

1. The specific situational contexts listed in the dictionary can be interpreted as triggers of stance-taking¹⁷;

2. the specific illocutive functions described in the dictionary refer to speech acts that can be attributed to the global speech act pattern ‘expressing one’s opinion’¹⁸.

The GEPHRI dictionary attributes non stare né in cielo né in terra to a ‘semantic field’ (“semantisches Feld”¹⁹) called “incoerenza” (‘incoherence / incoherency’), which in turn is attributed to a more global frame called “Menschliches Handeln und Verhalten” (‘human action and behaviour’), thus focusing on the ‘referential’ aspect of meaning (cf. 3.2 below). From the learners’ point of view, however, this choice may not be more useful than focusing on the pragmatic and discursive (i.e. communicative, cf. 2.3) aspects of the opinion frame. Either way, their understanding of this linguistic unit remains partial, unless they study the Phraseoframe of this construction in its entirety and thus acquire a holistic view²⁰.

The interconnectedness of semantic frames can be illustrated by the German FrameNet-Constructicon-project²¹, which aims at elaborating a constructicon of the German language comparable to the Berkeley FrameNet²²:

The ultimate goal of the German Constructicon […] is to identify and describe all constructions constituting the grammar of German in such a way that everything language users have to know in order to appropriately use and understand a construction is captured. Besides semantic, pragmatic, discourse-functional, and syntactic specifications, a full-fledged description of a construction also comprises information about relations to other constructions. Capturing the entire network of constructions in German constituting the constructicon is, to say the least, an ambitious long-term project […]. (Boas / Ziem 2018: 215)

Unlike the GEPHRI dictionary, the German FrameNet Constructicon does not specifically target foreign or L2 language learners and is conceived as a “non contrastive” constructicon (cf. Boas / Ziem 2018: 219)²³. Whilst partially subscribing to Croft’s point of view according to which constructions are language-specific and thus cross-language generalisations difficult to establish (cf. Croft 2001: 6; 283; Boas 2010: 5–7; Boas / Ziem 2018: 256), the methodology of the project is nevertheless partially contrastive insofar as the frame semantic part is built on the English model of the Berkeley FrameNet. Thus, the German opinion-frame (labelled “Meinung”) is more or less identical with the corresponding English frame, based on a theoretical framework developed by Fillmore (1982, 1985)²⁴. The definition of ‘frame’ implies a both intuitive and cognitive approach to meaning:

Frame Semantics offers an intuitive method of elaborating the analysis of form-meaning relationships by focusing on lexical semantic issues that are relevant to grammatical structure, among other things. It differs from other theories of lexical meaning in that it builds on common backgrounds of knowledge (semantic “frames”) against which the meanings of words are interpreted. (Boas 2010: 8)

The opinion frame of the Berkeley and the German FrameNet is structured around three core elements (CE) defining the meaning of the lexical unit (LU)²⁵ ‘opinion’, i.e. the CE ‘cognizer’, ‘opinion’ and ‘topic’: “A Cognizer holds a particular Opinion, which may be portrayed as being about a particular Topic” [terms in italics marked by colour code in the original]²⁶.

This knowledge-based definition of the opinion frame is of limited use for the language learner, since it does not take into account the (meta)discursive and pragmatic dimensions of the opinion frame in question. However, the establishment of systematic links to related semantic frames can give valuable insights into the interconnectedness between frames in terms of ‘families’ (“Framefamilien”²⁷), such as the connection with the frame ‘epistemic stance of the speaker’ (“epistemische Sprechereinstellung”²⁸). This frame refers to the LU in der Tat in terms of one of its ‘frame-evoking elements’ (FEE). They are categorised as frame elements signaling the epistemic status (“epistemischer Status”) of the propositional content of a phrase with respect to a ‘scale of certainty’²⁹. From the learner’s point of view, this definition provides insight into the meaning of the lexical multiword unit in der Tat merely to the extent that it operates on the level of stance-taking in a comparable way to other FEE listed in this context (i.e. gewiss, möglicherweise, scheinen, sicher, tatsächlich, vermutlich, wahrscheinlich, zweifelsohne). This, however, represents just one of the multiple aspects of meaning, namely what Lewis (1993: 84–85) calls the “modal meaning”. Other meaning aspects linked to the opinion frame can be found in different parts of the German FrameNet³⁰, but the complexity of the access structure makes it difficult for the learner to gather and process all the information needed to use expressions like in der Tat appropriately.

The difficulty of applying construction grammatical resources such as FrameNet to foreign language learning and teaching (FLL/FLT) has been recognized and discussed by a number of researchers within the recent field of Pedagogical Construction Grammar (PCxG – for an overview see Boas 2022). The German Frame-based Online Lexicon (G-FOL)³¹, a beginner learner’s dictionary of German for speakers of English (cf. e.g. Boas / Dux 2013; Boas et al. 2016; Boas 2022), represents an attempt to apply the “Seven Principles for Pedagogical Construction Grammar” by Herbst (2016) (cf. Boas 2022). For reasons of space, the following observations are restricted to the meaning-aspect regarding the frame-semantic dictionary entry “thinking: opinion”³².

The contrastive German – English presentation of the semantic frame, followed by a graphic representation of the abstract ideas of ‘thinking’ combined with ‘judging’ (= having an opinion) facilitates the learner’s access to the concept by applying the principles of a simultaneous bottom-up and top-down approach such as that recommended by Lewis (1993: 88):

In summary, meaning is made in three ways, by an implicit system of contrasts; bottom-up, a synthetic process of adding bits to each other, and by top-down interpretation, incorporating detail into an overall view based on previous, real-world (including language) knowledge. [Bold characters in the original]

According to Lewis (1993: 88), “meaning is a complex concept” which, since “the purpose of language is communication”, implies that “language teaching [and learning] activities should invariably be meaning-centred” (ibid.). Amongst the meaning aspects presented by Lewis (1993: 77–88) in his chapter “The Nature of Meaning”, the pragmatic and the negotiated dimensions are particularly relevant for FLL and FLT for at least two reasons:

(1) Expressing one’s opinion implies a methodology that interprets the meaning of specific linguistic (multiword) units in terms of evaluative speech acts produced with the intention of not merely sharing one’s point of view but also influencing the communication partner in some way. Hence, for the learner to be able to understand the meaning, they must become aware of the complex communication strategies underlying expressions of opinion by asking questions such as: ‘How is opinion expressed in this context and to what end?’ However, since the ways we communicate are culturally codified and the meaning of a vast number of linguistic units conventionalised, the learner will not be able to answer these questions without substantial help³³.

(2) Apparently banal and routinised expressions like meiner Meinung nach have specific meanings in different contexts. In the G-FOL, this expression is mentioned with respect to the particularity of its syntactic structure, but its meaning is not described separately from to the “noun opinion” that is generally “used like English opinion”³⁴. If the concept of ‘opinion’ is the same in German and English, this does not imply that the meaning of constructions containing the linguistic unit (LU) opinion can be fully understood by simply referring them to the cognitive opinion-frame. The following statement, found in the Gegenwartskorpora of the Digitales Wörterbuch der deutschen Sprache (DWDS³⁵), refers to meiner Meinung nach as a means to signal one’s readiness for negotiation:

Alles gilt als verhandelbar, wer nicht bereit ist zur Metareflexion, gilt als unempathisch, gar als gewaltvoll, wer ein Urteil fällt, wer nicht permanent darauf hinweist, er oder sie spreche jetzt aus rein subjektiver Perspektive, wer nicht jeden zweiten Satz mit ‘Meiner Meinung nach’ oder ‘Ich habe das Gefühl, dass’ einleitet, dem wird Nachhilfe in GFK [ = Gewaltfreie Kommunikation] verordnet.³⁶

For the learner, the communicative function of meiner Meinung nach is highly relevant if they seek to understand and use this expression appropriately. The lexical form of Meinung (opinion) determines the function of the construction to a lesser degree, since it is suggested that Ich habe das Gefühl, dass (‘I have the feeling that’) (see above) represents an equivalent and could thus be attributed to the same “family” of constructions.

In other contexts, however, meiner Meinung nach appears to function differently, as in the following (constructed) exchange:

• A: Ich bin ja so müde!

• B: Meiner Meinung nach solltest du dich ausruhen.

• A: Du hast gut reden! Erstmal muss ich mit der Arbeit fertig werden.³⁷

The sequential position and discursive context of the utterance suggest that B reacts to A’s complaint (i.e. “I’m so tired!”) by formulating a piece of advice (“If you ask me, you should have a rest”). Hence, the meaning of meiner Meinung nach is different in this configuration since it implies that the point of view of B contrasts with A’s point of view: B’s advice can be interpreted as suggesting that having a rest represents an advisable alternative to complaining about being tired. In this context, the English translation of meiner Meinung nach by if you ask me appears more appropriate than a translation containing the LU opinion³⁸. By associating meiner Meinung nach to a piece of advice, B emphasizes that his/her point of view is not simply different but also better adjusted to the situation, thus provoking A’s rejection (Easy for you to say!). It can therefore be argued that the opinion frame is connected to a ‘conflict’ frame which can be activated in various contexts and degrees, and be either moderated or emphasized (see above). This view is compatible with Lewis’ assumption of an ‘implicit system of contrasts’ as one of the three ways ‘meaning is made’: “Subconsciously we make meaning not only by recognising what the speaker has said, but by contrasting it with what has not been said” (Lewis 1993: 88; bold characters in the original).

Within a usage-based approach, it appears thus necessary to extend the ‘constructicon’ based on the Berkeley FrameNet in a way which takes into account the multiple aspects of meaning construction, namely by transposing the network of cognitive semantic frames into a multilevel model of meaning construction capable of representing expressions of opinion within a network of constructions connected to the opinion frame on at least two levels: (1) the level of (meta)discursive organisation associated to opinion-related speech act patterns (e.g. signaling openness to negotiation/ signaling a degree of certainty); (2) the level of interpersonal communication on which meaning emerges from constant negotiation (e.g. communicative strategies such as moderating or emphasizing conflictual elements linked to the opinion frame).

2.3 Comparing L1 and L2: some methodological issues

As discussed previously, errors regarding the German constructions meiner Meinung nach and meine Meinung ist, dass … have to be referred to processes of understanding (see 1., 2.1). The argument of insufficient understanding might be rejected by the learner on the grounds that they simply had not memorized the exact ‘form’ of each ‘word’ composing the expression. However, the number of errors linked to the functional-pragmatic aspect of these constructions in the written productions of my French learners’ GFL-corpus appears to be at least equal if not superior to the number of purely formal ‘mistakes’. If meine Meinung ist, dass had been understood and learned as a construction, the risk of morphosyntactic error might have been reduced. This leads us back to the question of ‘meaning’: What does a writer/speaker ‘mean’ when they say In my opinion?

In a phraseological perspective, EoO represent pre-constructed phrases whose predominant communicative function(s) must be described with respect to the contexts in which they occur. Taking Sinclair’s slogan “The phrase, the whole phrase, and nothing but the phrase” (Sinclair 2008: 407–410) a step further, I would like to argue that this type of expression can only be fully understood with respect to genre/text type as well as the particular situation in which it is employed. This top-down approach increases the importance of the pragmatic and interactional dimensions of the notion of meaning. According to Firth’s contextual meaning theory, meaning must be analysed “at all levels of linguistic analysis” (Firth 1957: 192). Describing context in the Firthian sense can provide valuable insight into a large variety of elements which are likely to influence the process of meaning construction and re-construction by L2 learners. One important aspect is the different situational and motivational background of L2 learners compared to native speakers: whereas students’ text productions are initiated by the teacher in an institutional (academic) setting and thus primarily extrinsically motivated, the native speaker’s productions are generally intrinsically motivated, namely in situations other than learning. This is to say that learners’ texts are focused at least partially on the language itself, being both a means of communication and an object of learning.

Within a contrastive approach, one question is whether part of the linguistic differences between L1 and L2 texts can be attributed to the double quality of L2 texts as both learners’ and learning texts (i.e. texts produced for learning). It could be argued that L1 and L2 texts are not comparable, since a number of accounted differences are directly imputable to the learning situation. The following table illustrates major differences between L1 and L2 text production situations with respect to the specific conditions on which the corpora of the present study are based:

A few criteria listed in this table represent simplifications in the sense that they emphasize the learning aspect of the L2 text production situation. The intention is to highlight differences which are likely to affect the writing style: since the writing task is generally imposed by the teacher in a learning situation, it is possible that a certain number of EoO are produced for the sole purpose of responding to the task by trying to meet the teacher’s expectations. Thus, the learner may express an opinion even though they might not actually have any to communicate on the given subject. This may lead to incoherencies and errors affecting all levels of text production, including constructions and formulations related to opinion. The question mark behind the teacher as addressee of the learner’s text suggests, however, that this might de facto be the case, but that this would not involve any substantial alterations in terms of ‘recipient design’. Most linguistic forms of ‘engagement’ of the reader rely on a series of highly conventionalised procedures and devices, such as direct address (reader pronouns, personal asides), appeals to shared knowledge, directives, and questions (cf. e.g. Hyland 2005: 182–186). It can be assumed that the concrete identity of the reader affects these procedures and devices only marginally. Another way of looking at this consists in considering language learning as a form of ‘game’ which enables the learner to acquire new skills and knowledge by way of playing. In this perspective, the learner ‘playfully’ constructs a reader whose attention is primordially directed to style and content³⁹ rather than lexical and grammatical correctness. Hence, the learner’s text can be considered globally comparable to the native speaker’s.

To conclude this theoretical section, I propose a summary of my basic postulates and assumptions:

1. ‘expressing one’s opinion’ represents a discursive frame within the ‘opinion frame’;

2. the frame of ‘expressing one’s opinion’ is deployed linguistically through a network of speech-act patterns;

3. these patterns refer to specific procedures on the semantic, morphosyntactic, pragmatic, interactional as well as the lexical and ‘formulation’ level. EoO are forms of formulaic language which are analysable as ‘constructions’;

4. semantic frames, patterns and grammatical constructions are interwoven, and constitutive elements of specific text patterns related to the opinion frame;

5. the structural, linguistic, and pragmatic properties of a text based on the opinion frame can be described as a complex argumentative text pattern;

6. text patterns are constitutive elements of genre and text type.

Examples of text types related to the opinion-frame are blog comments⁴⁰, radio/television comments, essay discussions, editorial, letter to the editor.

3.1 Data and method

The following exploratory study deals with one construction and two word forms: in der Tat, Meinung, and sagen. It is based on corpora compiled specifically for the purpose of examining the use of EoO by French learners of GFL and comparing them to the way they are used by German native speakers (GNS). The choice of in der Tat is motivated by the relatively high frequency in which it occurs in French learner’s text productions and my difficulties as their teacher in explaining why this expression is inappropriate or ‘unusual’ in most cases. The word forms Meinung and sagen were chosen as lexical elements frequently associated with the opinion-frame and thus likely to occur within speech-act-patterns related to ‘expressing one’s opinion’.

The general background of the study is as follows: the data of the GFL-corpus were collected over the past three years during my classes in front of French second, third and fourth year non specialist students at the University of Lorraine (France). All the data are written texts handed in electronically and responding to writing tasks explicitly involving the pattern ‘expressing one’s opinion’, typically elicited by formulations such as Kommentieren Sie … / Schreiben Sie einen Kommentar/Leserbrief. / Geben Sie Ihre Meinung ab!; Was ist Ihre Meinung dazu? The level of the students is relatively advanced and can be situated in a range between B1 and B2/ B2⁺ of the Common European Framework of Reference for Languages⁴¹. The GNS-corpus is compiled according to the criterion of comparability in terms of text type, but also in terms of length. Thus, since learners’ texts are generally short compared to GNS-texts, I restricted my choice of GNS-texts to the text type Leserbrief (‘letter to the editor’) in a wide sense: As these texts were published in online versions of German regional and national newspapers (e.g. Tagesspiegel, FAZ), it must be assumed that elements of the traditional text pattern have evolved or even disappeared, such as traditional introductory formula addressing the editors (Liebe Redaktion, …). The degree of comparability⁴² of the GFL- and the GNS-corpus is relatively high with respect to the speech-act-pattern ‘expressing one’s opinion’, but it remains ultimately insufficient regarding specific text types such as ‘letter to the editor’. Nevertheless, my French learner’s corpus contains 23 documents explicitly referring to this text type, thus increasing the chances of finding comparable text- and speech-act-patterns in both corpora. I have chosen not to include a French native speaker’s (FNS) corpus, considering the present study as simply a first step to improving French learners’ writing skills. Its focus lies on the revelation of culture and language specific patterns and constructions of ‘expressing one’s opinion’. Only in a second step can we attempt to describe these patterns and constructions in more detail.

The table below provides a general overview of the data. The GFL-corpus is divided into two sub-corpora, whereas the GNS-corpus, solely based on the text type ‘letter to the editor’, does not require any subdivision.

Method: The corpora were compiled in Sketch Engine (cf. Kilgariff et al. 2014) in order to enable basic queries concerning frequencies and concordances. The latter were extracted for a qualitative analysis in order to determine the underlying constructions and patterns of a complex argumentative structure. Larger parts of texts were extracted and annotated manually according to the type of speech acts performed within argumentative discourse. The descriptive categories were adapted, and partially altered, from typologies developed by Tutin (2019) with respect to ‘prefabricated sentences’ used in interaction⁴³ as well as from Granger / Paquot (2008) regarding phraseological categories for textual and communicative phrasemes.

3.2 First results

In the following, I will start by presenting general findings concerning the use of the word Meinung, before presenting the construction in der Tat by combining empirical results with a more detailed qualitative analysis. Finally, I will present a few interesting findings with respect to sagen-constructions.

The choice of Meinung was motivated by my expectation of finding a relatively high number of occurrences in both the GFL and the GNS productions. Surprisingly, only two occurrences could be found in the GNS productions, whereas the GFL-corpus contains 47 occurrences⁴⁴. The GFL-subcorpus based on the text type ‘letter to the editor’ contains 7 occurrences and thus distinctly more than the GNS corpus, even though the latter is more than three times bigger (16,600 words compared to 6,210 words). The qualitative analysis revealed, moreover, that both occurrences in the GNS-corpus are used in descriptive contexts unrelated to the specific metadiscursive and pragmatic functions of ‘expressing one’s opinion’. In the French learners’ productions, on the other hand, 33 occurrences out of 47 can be attributed to this type of speech act. Their specific functions can be described in terms of ‘(meta-)attitudinal’⁴⁵ constructions such as meiner Meinung nach. It can be noted that they are frequently connected to metadiscursive acts of discourse organisation, as in the following examples:

The combination of metadiscursive devices of text organisation, such as endlich (1) and sehe ich ein weiteres (…) Argument (2) with Meinung-constructions creates an effect of over-emphasis on interactive steering and coherence-marking. This appears relatively unusual to a German native speaker, an impression requiring verification on bigger GNS-corpora. Finally, the relatively frequent cases of unusual constructions are less often due to incorrect syntax than to an incoherent combination of constructions. The following examples illustrate the sometimes comical effect:

The pragmatic-semantic incongruence created by these sentences is based on a confusion of collocational relations: the adjective gemischt is usually part of the collocation ein gemischtes Gefühl/ einen gemischten Eindruck haben (E: ‘having mixed feelings / impressions’), its use being restricted to the psychological aspect of perception. In the second example, the comical effect relies partially on the restriction of the expression seine Meinung teilen (E: to share one’s opinion) to humans, thus bringing the “Text” literally to life. These types of error indicate a certain degree of awareness with regard to constructions, despite the learners’ insufficient linguistic competencies concerning the lexico-grammatical properties of this construction.

The EoO in der Tat is used very frequently by French learners of German. Their considerable difficulties in using this expression appropriately is linked to its syntactic, semantic and pragmatic fixity and the existence of a French translation which is, however, only partially equivalent (F: en effet; E: indeed). The corpus-based study of this expression leads to similar results compared to Meinung in terms of frequency: With only two occurrences in the GNS-corpus its number is significantly lower compared to the 36 occurrences in the GFL-corpora. The specific discursive function of the two occurrences in the GNS-corpus can be glossed as marking the evidence of an assertion within argumentative discourse:

This example of argumentative discourse deploys a series of interlinked speech acts as part of a textual pattern which can be analysed according to the global speech act pattern of ‘evaluating linguistically’ (G: “sprachliches Bewerten”) developed in a text linguistic perspective (see e.g. Sandig 1986; Trevisan 2013). The specific function of the expression in der Tat within this example can be referred to a global argumentative pattern inside of which it carries out the function of an operator by marking the evidence of the assertion to which it is associated. The meaning of the assertion in question can be glossed as follows: The traditional family (“klassische Familie”) is an undeniable fact, i.e. “not an invention of an ideological school of thought”. In addition to what Hyland / Wang / Jiang (2022) call the ‘interactional’ function (see section 2.1), in der Tat can be seen as an ‘interactive’ device preparing the thematic development of the following sentence. On the one hand, in der Tat represents a ‘warrant’ to the ‘claim’⁴⁷ by presenting the existence of extended families in the past as a proof for the claim’s veracity. On the other hand, the salience of the sequential position of in der Tat provides emphasis and can be interpreted as an interactive signal regarding the relevance of the claim. Since the relevance is not (yet) explicit, in der Tat informs the reader implicitly that a proof of the claim will follow: Es gab zum Beispiel … (‘For instance, there was …’). Since a granular analysis of this cannot be fully provided in the present article, this brief account can only provide first insights into the complexity of EoO.

The fact that only 20% of the EoO produced by French learners are used appropriately can serve as an illustration. A closer analysis of the contexts in which inappropriate use occurs shows a tendency of French learners to employ them as textual coherence markers. For reasons linked to the length of the present article, I will restrict my observations to one example:

In this context, the expression is used by the French learner to signal the thematic link between the assertion produced in the previous sentence and the illustration in the second. Since the thematic development provided by this second sentence is incompatible with the evidence-marking function discussed above, the use of in der Tat must be rendered inappropriate. From the point of view of a GNS, there does not seem to be any better way of indicating the intended meaning than by simply eliminating the expression. Hence, it is imperative for French learners to study the complex discursive functions of in der Tat in the contexts in which they are employed by GNS.

The significance of the results obtained from the quantitative and qualitative analysis of the data will have to be confirmed by a contrastive analysis by integrating a comparable FNS corpus into the study.

The findings with respect to constructions occurring in combination with the word form sagen will be restricted to a few essential observations. The choice of sagen was based on the hypothesis that verba dicendi are frequently associated with the EoO-pattern, since their metadiscursive dimension makes them a valuable resource for stance-taking purposes. An example of a stance-taking construction containing the word form sagen is the idiomatic speech act Na sag mal! (E: ‘You don’t say! / I can’t believe it! / Wow, that’s impressive / …’)⁴⁹. Whereas this construction typically occurs in discourse-in-interaction, formulations in written productions based on the text type ‘letter to the editor’ tend to be more descriptive regarding the author’s sentiments. This assumption is corroborated by my findings with respect to EoO in the GNS-corpus. The morphosyntactic analysis of the GNS-constructions reflects a preference for impersonal formulations compared to the EoO found in the GFL-corpus.

The following table gives an empirical account of sagen-constructions in the GNS-corpus and the French learner subcorpus based on the text type ‘letter to the editor’, using Sketch Engine.

Even though the limited number of our data does not enable us to draw any general conclusions, our analyses have shown that there may be a stronger tendency for French native speakers to use sagen-constructions compared to German native speakers. This might be ultimately an indicator of cultural and linguistic differences with respect to speech and text act patterns in the two languages in question, an observation requiring further study.

The proportion of constructions linked to the EoO-pattern with respect to the total number of ‘more or less fixed multiword units’⁵⁰ analysed is roughly the same in both corpora. The qualitative analysis was carried out in three steps: (1) filtering of communicative constructions; (2) general morpho-syntactic description of each construction; (3) in-depth analysis of the specific discursive and pragmatic functions of each construction regarding the specific contexts.

Step (1) consisted in a simple division of all occurrences into two phraseological categories based on a rough distinction between ‘referential’ (i.e. descriptive) function on the one hand and the ‘textual’ / ‘communicative’ function on the other, thereby following the extended version of Burger’s typology (1998) presented by Granger / Paquot (2008). In this extended version, Burger’s distinction between the ‘referential’, ‘communicative’ and ‘structural’ function is modified insofar as the ‘structural’ function is renamed ‘textual’ function, thus extending the number of phrasemes which could be potentially attributed to this category. The extension of the analytical scope corresponds to a text analytical approach going beyond the level of the phrase and the sentence (see above, 2.2). For the purpose of the present study, however, I did not consider it useful to establish a clear categorial distinction between the communicative and the textual function, since EoO are primarily seen as metadiscursive resources whose interactive (i.e. textual) and interactional (i.e. communicative) functions cannot always be clearly separated (see above).

The following examples taken from the GNS-corpus illustrate the basic distinction between referential and communicative functions of phrases containing sagen:

Despite the argumentative context in example (9), the sentence based on the temporal/ iterative structure Wenn (ich) …, dann (sage ich)…, is purely descriptive: in referring to the writer’s habit of calling it a day after eight hours of work, the VP ich sage “Feierabend” fulfills a referential function. Unlike in example (10), sagen is not part of a construction in the sense defined above (see introduction)⁵¹.

The construction Das musste einfach mal gesagt werden in (10), on the other hand, carries out a communicative function since the meaning of sagen in this context is inextricably linked to the meaning of the other components and thus represents a ‘holistic sign’ (see introduction). Its function in the present context is both to mark the end of a sequence and to emphasize the relevance of what is said, thus combining interactive and interactional aspects of metadiscourse.

The lists below provide an account of all EoO-constructions found in the GFL and GNS corpora, followed by a schematic overview regarding their morpho-syntactic and functional-pragmatic properties.