Italian ma ‘but’ in deverbal pragmatic markers: Forms, functions, and productivity of a pragma-dyad

This paper focuses on the development, paradigmaticization, and productivity of a set of complex pragmatic markers in Italian, which are constituted by two distinct elements, namely the adversative conjunction ma ‘but’ and a deverbal pragmatic marker (e.g., ma dai ‘come on! really!’, literally: ‘but give’, or ma piantala ‘just stop! give it a rest!’, literally ‘but dump it’). The main idea we will develop is that such a complex pattern can be better described in terms of a pragma-dyad, i.e., a dyadic construction with a pragmatic meaning, featuring a fixed element which systematically combines with a set of preferential fillers. In our case, the fixed element is ma, which generally signals a contrast with the interlocutors’ point of view, thus shaping the pragmatic meaning of the resulting complex in terms of interactional contrast. Such a meaning is then functionally enriched through a variety of fillers compatible with the schema, which actualize it in conveying mock politeness, disagreement, counter-expectation, and pragmatically neighbouring values. By providing a corpus-based study of the development and productivity of these complex markers, we illustrate the empirical and theoretical advantages which a pragma-dyadic approach can offer in exploring processes of functional enrichment involving complex markers.


Introduction
In this paper 3 we provide a functional account of a set of complex Pragmatic Markers (henceforth, PMs) widely attested in in present-day Italian (PDI) which are all made up of two elements, namely the adversative conjunction ma 'but' followed by a PM made up of a pragmaticalized verb in the 2 nd -person singular of the imperative mood. The set of verbs functioning as possible fillers in this specific pattern is listed under (1). 4 (1) ma scusa lit. 'but excuse IMP.2SG ' ma dai lit. 'but give IMP.2SG ' ma piantala 'but stop IMP.2SG ' ma va' lit. 'but go IMP.2SG ' ma vieni lit. 'but come IMP.2SG ' These six PMs result from the systematic combination of ma plus an independent PM already existing in the language, with the exception of vieni, which cannot be used in isolation with a pragmatic meaning, as we will see in detail in § 3.4. On the whole, these markers are pervasively used in PDI and, given their frequency and productivity, we believe they deserve a dedicated study.
The motivation behind this study also rests on the fact that in the past few years only little attention has been paid to the dynamics through which recurrent associations of discourse or pragmatic markers emerge and lexicalize in a given language. Pons (2018), who represents a notable exception to this trend, notes that this topic «is perhaps the only point in which the (in other respects) large amount of research on this subject has not yet produced relevant contributions». His paper discusses Spanish spoken data through the lens of the Val.Es.Co. model of discourse segmentation and makes some useful remarks to qualify the syntagmatic relation holding between two 3 Although the paper has been conceived by the two authors together, Sections 1, 2, and 3.1-3.2 have been written by Piera Molinelli and Sections 3.3-3.5, 4 and 5 by Chiara Fedriani. A preliminary version of this study was presented at the iMean Conference (Bristol, April 6-8, 2017) and at the International Conference on Discourse Markers in Romance Languages 5 (Louvain-la-Neuve, November 8-10, 2017): we thank participants at the Conference for their useful comments. We are also grateful to Karin Aijmer, Kate Beeching, and Chiara Ghezzi for their insightful remarks and suggestions. 4 It has to be noted here that this pattern also includes other possible fillers, such as verbs in the subjunctive (ma andiamo 'lit. but let's go', ma finiamola 'lit. but let's stop it') and politeness markers (ma per favore 'but please', see Fedriani 2019). They will be briefly addressed in the discussion developed in Section 4. elements in a pragmatic combination, distinguishing between adjacency of neighbouring markers and combination proper. Another relevant study is that by Fraser (2013), who takes into account possible combinations of contrastive Discourse Markers in English, identifying semantic and formal restrictions in their mutual association. These are both valuable contributions, which look at possible combinations between virtually all the Discourse Markers included in the corpus (Pons 2018) or, within a narrower perspective, between contrastive markers of all kinds (but, however, yet, on the other hand, among many others, see Fraser 2013). In this paper, by contrast, we focus on a different case, that of one specific pattern consisting of a fixed element which systematically combines with a circumscribed set of formally defined fillers. This is an important point, because, to our knowledge, there are no studies offering a principled account of pragmatic complex patterns of this type.
At a more general level, the collostructional approach (Stefanowitsch / Gries 2003) has worked out a method of corpus-based linguistic inquiry which investigates statistically the degree of attraction (or 'collocational strength ', cf. Wiechmann 2008) of words in a collostruction. However, the collostructional method does not seem particularly suitable in the realm of pragmatics, above all due to a crucial difficulty in automatically recognizing pragmaticalized vs. literal uses of a given item -a distinction which, on the other hand, requires manual introspection (as we will see in § 2).
Lastly, the available pool of data and discussion only contains cursory information from Italian. In particular, the syntagmatic contiguity of pragmatic elements has been described by Bazzanella (2001: 44) in terms of cumuli 'accumulations', when two or more markers with the same function are juxtaposed to each other, and of catene 'chains' if they instead have different values. However, this description does not attempt to capture the dynamics whereby recurrent elements systematically combine to give rise to further complex markers with their own individual pragmatic function.
Therefore, this study aims to fill several gaps. It does so by suggesting a theoretical notion to look at complex patterns developing pragmatic meaning, at the same time as providing a corpus-based, functional description of such complex units in Italian by focusing on the productive schema featuring ma and a PM in the imperative.
In the complex items we are looking at, the adversative conjunction ma 'but' is an invariable component which is present in all the resulting markers, and which we call co-unit. The set of possible PMs compatible with ma can, in turn, vary: We call these different fillers pragma-units (guarda, scusa, dai, and so on). We employed the prefix coin the label co-unit to highlight the functional and semantic co-operation performed by this element when interacting with its possible fillers, and to point to its role in co-constructing of procedural meaning. We suggest using the label of pragma-dyad for the complex PM resulting from a co-unit and a pragma-unit. Within a given pragma-dyad, the co-unit is fixed: In our case, its slot is always filled by ma, which substantially contributes to the resulting meaning of the pattern in terms of contrast. The specific pragmatic value of each instantiation of the pragma-dyad is then defined by the different pragma-units: scusa 'excuse (me)' has a very different pragmatic meaning from, for example, guarda 'look'. Taken together, the co-unit ma and the pragma-units it can combine with give rise to a dyadic construction with a pragmatic meaning (Figure 1). In our view, the notion of pragma-dyad has the advantage of accounting for complex linguistic units which result from the recurrent association of two autonomous elements with two different meanings M 1 and M 2 , both existing as independent units in a given language but that, combined together, give rise to a more abstract template with a new meaning M 3 , built on, but different from, M 1 and M 2 . We examine this point in more depth, looking at the way in which the co-unit and the pragma-units contribute to the resulting general meaning M 3 of the pragma-dyad.
In Italian, ma has two main values connected to the functional core of adversative contrast, namely a totally adversative meaning, with an oppositive value ('not A, but rather B', with B substituting A, see ex. 2), and a partially adversative meaning, with a corrective value deleting an inference ('not A, but B', with B limiting or clarifying A; cf., e.g., Dardano / Trifone 1995: 444). When bearing the partial adversative meaning, ma can express counter-expectation (ex. 3) or evaluation (ex. 4): (2) Non è inglese ma francese 'He's not English but French' (3) Non è alto, ma gioca a basket 'He's not tall but he plays basketball' (4) Il libro è lungo ma piacevole 'The book is long but nice' As we have just seen, therefore, typically ma licenses deletion or correction of inferences, thus working as a multifunctional 'inferential operator' at the level of both discourse management and discourse structuring. At an interactional level, ma can be used as a turn-taking or turn-leaving device, and this correlates with a preferential and frequent sentence-initial position (Molinelli 2010: 262). Moreover, it can also function as a topic-change marker: the interlocutor expects the ongoing topic to be continued, but the speaker changes it. An example of this function is given in (5), where the first speaker indicated the starting point of a path and is ready to move on and explain the route, whereas his interlocutor drives the conversation back to the prior topic to elaborate it further, signalling the topic-shift by prefacing it with ma, which points up thematic contrast: (5) p1G: la partenza è# una croce e da lì parti e c'è tutto il percorso che poi<ii> mo te dico ti dico# p2F: ma la croce sta sul televisore# o sulla macchina? (CLIPS Corpus, DG-mtA01R) 'p1G: The starting point is a cross and you depart from there and there is the whole route that then now I tell you I tell you p2F: but is the cross on the TV or on the car?' Along similar lines, ma can also signal counter-expectation, in terms of an unexpected contrast with the content conveyed by the interlocutor (see Borreguero 2009). As we will see in the course of this paper, this value is frequently activated within the pragma-dyad we are looking at.
Summing up, ma has a clear inter-clausal procedural function, providing the interlocutor with instructions as to how to interpret the link between two clauses -generally speaking, with the meaning of contrast (Mauri / Giacalone 2012: 192). In our pragma-dyad, ma retains a metonymic 'hint' of its adversative nature (Beeching 2007), pointing to a conflict between two contrasting views at the broader level of discourse. At the discourse level, thus, ma substantially shapes the resulting meaning of the pragma-dyad by building upon its «semantic potential» (in the sense of Norén / Linell 2007), which revolves around the core value of interactional contrast. As we will show, such a semantic core can be syntagmatically redefined with related, but different values when combined with different pragma-units. In other words, the specific way in which such a meaning of contrast is then actualized with different semantic-pragmatic nuances is constrained by the original lexical content of the various pragma-units ma co-occurs with.
Since, as we have just pointed out, the pragma-dyad is associated with a given meaning, we interpret it as a specific type of pragmatic construction, that is, a form-function pairing made up of a formal component (in this case, the co-unit ma followed by a set of possible PMs which are compatible with it), and a pragmatic function performed globally by the resulting complex (i.e., that of interactional contrast). Figure 2 represents the form-function pairing of the pragma-dyad under scrutiny. Within the constructionist perspective, particularly relevant to our study is the constructionist usage-based view of productivity, according to which productivity is the ability of a pattern to extend its structure to other types, or, from a different perspective, the likelihood of encountering new items entering a construction over time (Bybee / Thompson 1997). Productivity has thus been understood in terms of constructional extensibility (Barðdal 2008). In the above-mentioned works, this notion has been worked out primarily on syntax, notable exceptions being Fried / Östman (2005), Fischer (2010, and Fischer / Alm (2013), who provided functional accounts of different PMs in constructional terms. This paper takes its inspiration precisely from this line of research, although it depends on a less radical view than the general Construction Grammar-sense. What we have done is to take some notions drawn from a constructionist approach to explore the combinatorial potential of a co-unit with its pragma-units and the formal and functional relations holding between the two. Some notions gleaned from this framework seemed thus appropriate to the investigation of the emergence and spread of our pragma-dyad, possible extensions to new deverbal PMs, and the semantic and pragmatic factors which determined the recurrent combination of ma and its lexical fillers. Previous approaches also shed light on processes such as mutual influence and analogy (De Smet / Fischer 2017), formal resemblance (Octavio de Toledo 2018), and functional inheritance: we will take these processes into consideration in the discussion in Section 4.
Building on these premises, the remainder of this paper is organized as follows. First, a description of the corpus we have used, and some methodological assumptions are in order ( § 2). In § 3 we provide a functional description of the ma 'but' + pm in the imperative pragma-dyad in PDI, focusing on its different outcomes and also providing some contrastive remarks. In doing this, we will first briefly describe the meaning(s) of each pragmaticalized verb when it functions as an autonomous PM (this is the case of scusa: § 3.1, guarda: § 3.2, dai: § 3.3; and piantala: § 3.4), and the specific values they subsequently acquired when used in combination with the co-unit ma. We then turn to the pragmatic development undergone by va' and vieni ( § 3.4), two motion verbs which are less pragmaticalized in PDI. In Section 4 we discuss the process of paradigmaticization undergone by the pragma-dyad and its productivity. Section 5 concludes with a summary of the analysis developed in this paper, reviews the results, and assesses the implications for the adoption of the notion of pragma-dyad to research in the field of pragmatics.

Corpus and data
The analysis is based on the itTenTen corpus, an Italian web corpus part of the TenTen corpus family created through web crawling, a software designed to explore the web, analyzing, filtering, and copying texts from the Internet. We used the itTenT-en10 version, which contains a total of 2.5 billion words. Given the nature of the data considered, which mostly come from chats, blogs and forums, a few words on the characteristics of electronic discourse are in order.
As it is known, digital interactions are typically lacking in social, relational and affective richness relative to traditional forms of communication (see, e.g., Kalman / Gergle 2010), and this may correlate with a greater use of explicit linguistic strategies expressing both positive and negative attitudes, also including, as we will see, graphical features. In the passages commented on in this paper, the most common graphical strategies adopted in this context are repetitions of letters and punctuation marks, and the use of capital letters, which are strategically exploited as cues with which the digital writer reproduces constitutive traits of spoken non-verbal communication. Since these non-standard behaviours convey linguistic and attitudinal significance, we have of course reproduced them precisely as they were written in the original source -also containing, in some cases, some grammatical mistakes.
Interestingly, a further feature to underline here is that in digital interaction users can communicate anonymously and are therefore less committed to the potential negative implications of violations of 'politic' behaviour and politeness norms. As a result, linguistic creativity and expressivity is more frequently admitted or even enhanced, because it is seen as a creative mode of entertainment, exploiting the ludic component of a virtual environment where basically no risks are involved (see e.g. Leech 2014: 235). As we will see, the specific nature of digital communication and its inner potentialities has probably fostered the massive use of the complex PMs we are going to explore in the next section, which in many cases express face-threatening, or even impolite, functions.
Lastly, it has to be underscored that in order to disambiguate literal and pragmaticalized uses of the deverbal markers under scrutiny a close manual introspection of all tokens was needed. This is particularly true in the case of ma va' 'come on!', often spelled wrong in our corpus of web texts, i.e., without the apostrophe (ma va 'but (s/ he) goes'). As a result, the PM became homographic with the present indicative 3 rd -singular form va '(s/he) goes', giving rise to many cases of ambiguity which required a closer inspection.
'Sorry coach for my absence on Monday 6 April' Its apologetic meaning is usually employed to introduce a secondary action which may be interpreted as a Face Threatening Act (FTA), and this constitutes the majority of its uses: scusa thus functions as an alerter which 'warns' the interlocutor against what follows (Leech 2014: 122). In this case, scusa can have three functions depending on the type of secondary act it has scope over. First, it can be used to remedy a past or immediately forthcoming breach of 'etiquette' or other minor offences on the part of the speaker, for instance interrupting, mitigating a request, refusing, or, as in ex. (7), asking too many questions: (7) Dov'è che ti sposi? Dove fai il ricevimento? Ti sto facendo un po' di domande, scusa … 'Where are you going to get married? Where are you holding the reception? I am asking a lot of questions, excuse (me)' Second, scusa can be inserted before a strong criticism or disagreement, as a way of mitigating it (ex. 8): (8) Scusa, ma permettimi già di dissentire. Come dicevo poco fa il romanzo, secondo me manca di tutta una parte di concretezza e realtà 'Excuse (me), but allow me to disagree. As I was saying, in my opinion the novel lacks concreteness and reality' Last, it can also function at the level of discourse organization, for example to take the turn or ask a question, catching the interlocutor's attention (ex. 9) or taking time for online planning or to correct oneself (ex. 10). In such cases, scusa still behaves as a formulaic apology repairing minimal offences.
(9) Scusa un'altra cosa, riguardo il sacco a pelo, siccome ho un bagaglio a mano e rischio che col sacco a pelo diventi troppo grande, mi chiedevo: "ma veramente la notte fa così freddo lì. 'Excuse (me) one more thing, about the sleeping bag, since I'm taking hand luggage and there's the risk it might become too big with the sleeping bag, I was wondering: "but is it really that cold at night over there?' (10) Oppure sta' fuori, tanto la principessa è abbastanza "piccola" da passare dalla porta dorata, scusa intendevo magra. 'Otherwise she stays outside, since the princess is 'small' enough to pass through the golden door, pardon me, I meant thin enough' In sum, the pragmatic functions of scusa cluster around the redressive value which is implicit in the lexical meaning of the verb. Its uses position the speaker as 'feeling bad' about an action and wishing to repair a 'potential' damage to interpersonal relations with an interlocutor (Ghezzi / Molinelli 2019).
Turning now to the ma scusa pragma-dyad, it should be noted that it is far less frequent than scusa: we have counted 41,832 occurrences of scusa and 2,038 attestations of ma scusa in our corpus. Its lesser frequency is only to be expected, since scusa is the apology formula par excellence in Italian and is widely employed as a routinized marker in everyday interactions. Ma scusa, by contrast, has developed a more specific pragmatic function, which is rather oriented towards the pole of interactional impoliteness. Let us describe its functional spectrum in some detail.
First, ma scusa can express face attack apologies, thus introducing a potentially FTA for which the speaker expresses justification. As Ghezzi / Molinelli (2019) note, these uses «cannot be considered impolite, as the polite move, i.e. the apology itself, encompasses the impolite move, e.g. the criticism, and thus there is no mismatch of im/politeness». In (11), for example, the speaker prefaces an objection with ma scusa, and in (12) uses it before a suggestion: the insertion of ma scusa is needed since s/he presumably feels that the content of his/her secondary act may be interpreted as intrusive or inappropriate by the interlocutor. In both cases, scusa still carries a genuine politeness value, but ma enriches the apologetic meaning with an additional nuance of interactional contrast: in (11), the speaker suggests an adversative point of view which may be in contrast with his interlocutor's; in (12), the contrast stems from a potential attitude clash between the speaker's proposal (making tamarillo jam) and the interlocutor's reaction. Interestingly, in this latter case ma scusa is orthographically isolated from the secondary act it has scope over through a semi-colon -this presumably corresponding to a 'comma intonation' which is typical of highly pragmaticalized units: (11) ma scusa perche non chiami F. e gli chiedi un parere 'but excuse (me), why don't you call F. and ask her opinion' (12) Ma mg, non è possibile!!!! Hai anche il tamarillo!!! Ma scusa: tu che sei il guru delle marmellate…una marmellata di tamarilli no??? 'But mg, that's not possible!!! You also have tamarillo!!! But excuse (me): you're the jam guru…why don't you make tamarillo jam???' In our view, cases like (11) and (12) constitute the bridging contexts leading to the conventionalization of an impolite meaning developed by ma scusa. In the majority of cases, indeed, ma scusa gives rise to a mock politeness message, i.e., a message which has «an impoliteness understanding that does not match the surface form or semantics of the utterance or the symbolic meaning of the behaviour» (Culpeper 2011: 17), and this triggers an implication of impoliteness. Thus, this pragma-dyad frequently introduces criticisms or challenging questions, as in (13), where the speaker refuses to do what the interlocutor asked. In such a context, ma scusa is not used as a sincere apologetic formula but serves to emphasize a move of disagreement. Ma scusa is also used to preface rhetorical questions that have a sarcastic value, as in (14). We also found one case where the rhetorical question introduced by ma scusa is a conventionalized insult (ex. 15). In all these contexts the impolite reading of scusa is triggered and reinforced by the prefacing ma, which exacerbates the overall conflictive nuance and inviting the interpretation of scusa as a mock apology.
'but excuse (me), do you think you can say things like that?' (15) ma scusa, chi ti credi di essere? se una persona nn sta bene tu le dici così? vorrei proprio vedere te nei suoi panni!! 'but excuse (me), who do you think you are? If a person is not feeling well you say that? I would really like to see you in her shoes!!' The above examples clearly show that the meaning of interactional contrast conveyed by ma is crucial in shaping the mock apologizing value of scusa and, more generally, in characterizing the pragma-dyad as an interactional move of opposition introducing a secondary action constituted by an intentional FTA.
In the light of this discussion, hence, ma scusa can thus be broken down as follows. While scusa alone can be paraphrased as "I'm sorry that I'm going to say something unpleasant to you" and its effect at most is to mitigate slightly the impoliteness of the face-threatening act, the addition of ma is significant in encouraging the impolite interpretation and can be explicated as follows: "I'm sorry, but despite that, I'm going to say the unwelcome thing that I have to say" (cf. Leech 2014). It is now worth noting that it is precisely such a conflictive use which predominates in the first 100 hits of ma scusa in our corpus (see Table 1). Summing up, we have seen that the adversative meaning of the co-unit ma generally helps to signal a contrast with the interlocutor's point of view within the pragma-dyad. When added to the pragma-unit scusa, it develops a more specific meaning: it enables the speaker to pretend to mitigate a potentially FTA, thus shaping the pragmatic function of this specific pragma-dyad in terms of mock politeness, as illustrated in Figure 3. Therefore, the general meaning of interactional contrast conveyed by the abstract template is actualized in terms of mock politeness in the specific instantiation realized by ma scusa.

Ma dai
This pragma-dyad is the outcome of the recurrent association of ma and dai (literally, 'give (you)!'), a pragmaticalized form of the 2 nd -person singular imperative of dare 'to give', meaning 'come on!, all right!, really? no way!'. Note that dai is highly pragmaticalized in PDI and not perceived as being related to the verbal paradigm of dare by speakers. Further evidence for its high degree of fixation also comes from the fact that it lost the possibility of number agreement with plural interlocutors, as shown in ex. (16). In this respect, dai differs from scusa, which is semantically transparent and allows modification in its inflection (scusate, mi scusi).
Trying to explain the motivations behind the pragmaticalization of this verb, however, Fedriani / Ghezzi (2014) have suggested that they rest on the fact that dare is a verb of exchange par excellence, and this semantic component played an important role in triggering the development of its pragmatic meaning. Verbs of giving do tend to imply a transfer of material, hence a negotiation, and this results in its deep embedding in the basic canonical schema of interactional verbal exchange. Moreover, the act of giving projects a privileged focus onto the transferred item (giving something), thus triggering a metaphorical implicature according to which propositional contents are conceived as concrete entities transferred from the speaker to the hearer in the dialogic space in terms of textual "objects". A brief functional description of dai will shed light on this issue.
First, dai can function as an exhortative PM used by the speaker to ask the interlocutor to give the requested action to him. In this case, the request is metaphorically conceived as a concrete object that can be given, taken, and exchanged, as in (16), where the speaker metaphorically asks the interlocutors to 'give' him the action of using something: (16) <utilizzatela / dai /ragazzi> (C-Oral-Rom corpus, ifamcv02) 'Use it, come on, guys' Dai can also be used as a marker of agreement: in this case, the speaker asks the interlocutor to "give his assent", as in (17), also with a leave-taking function, since coming to an agreement is an essential precondition for a smooth conversational closure (ex. 18).
(17) però è bellina/come idea /<dai> (C-Oral-Rom corpus, ifamcv26) 'but it's not bas as an idea, let's do it' In sum, the semantic import of the original lexical semantics of dare sheds light on the transferred entity. Within the interactional exchange, this enhanced the pragmaticalization of dai to express one's stance with regard to the reception of a given communicative object. Building on these pragmatic functions, let us now see what additional meanings dai developed when used in systematic combination with ma.
The ma dai pragma-dyad has two core values, which constitute pragmatic elaborations of typical functions performed by dai with the addition of the feature of contrast: an exhortative value, to convey an encouragement in opposition with the interlocutor's point of view, and a mirative value, expressing counter-expectation.
First, ma dai can be use similarly to the exhortative dai, but with an adversative nuance, since the exhortation is in clear opposition and contrast with the interlocutor's point of view. In (20), for instance, the speaker tells the story of a lunch during a holiday in the Mexican city of Merida, when his wife showed clear signs of aversion to Mexican food, longing to have macaroni instead. As a result, he exhorts his reluctant travelling companion to adapt herself to the local cuisine, reinforcing his invitation by prefacing it with ma dai, which serves to signal a clear contrast to his wife's attitude. When introducing an opposing view, ma dai can also serve to strengthen the speaker's perspective, especially while trying to convince the interlocutor, as in (21), where Ester attempts to persuade Caterozza that Bill's behaviour is not a nightmare, but instead is obvious (contrast), and implicitly exhorts her to 'take on' this perspective: The value of contrastive exhortation conveyed by ma dai in contexts such as those given in (20) and (21) can be interpreted through the lens of the notion of procatalepsis discussed by Beeching (2009): the speaker acknowledges that there is another point of view, but through this concession s/he then strengthen her own argument. As Beeching (2009: 82; bold original) convincingly claims, «by downtoning the force of an assertion and conceding part of the argument, the speaker is paradoxically able to be more convincing and boost the argument». We can thus conclude that in such contexts the adversative semantics of ma elaborates in terms of contrast the speaker's point of view, thus boosting the exhortative meaning conveyed by dai while inviting the interlocutor to adhere to the speaker's alternative perspective, as exemplified in Figure 4.  Interestingly, building on this function the pragma-dyad further developed an interjection of overt contrast, signalling impatience and introducing a move of strong disagreement. In our corpus of online digitized Italian, ma dai is used in these contexts as an interjection-like item: it usually prefaces a move of dissent with specific orthographical correlates, being frequently followed by punctuation marks which separate it from the following clause (ex. 22-23). When ma dai has such an exclamatory import it can also be added as a sort of afterthought signalling dissent, as in (24), where it is placed at the end of a polemic statement to express total disagreement, in its univerbated form maddai (see below).
(22) Lo scooter non è sinonimo di moto, ma sono due oggetti distinti. Sarebbe come dire che il quad è un automobile perchè ha 4 ruote… ma dai! 'A scooter is not synonymous with a motorbike, they're two distinct objects. That's like saying that a quad is a car because it has four wheels… come on!' In cases such as the ones given in (22) to (24), dai is partially bleached of its lexically-driven exhortative meaning and functions instead as an almost void pragma-unit, routinized in combination with ma without semantically contributing to the resulting pattern. Therefore, in exclamatory contexts ma dai is a frozen interjection with a holophrastic value, where the pragmatic meaning is actively construed mostly by the adversative value of ma, thus pointing to a strong contrast with the interlocutor's point of view, as summarized in Figure 5. The second core value enacted by ma dai is the expression of counter-expectation and surprise (ex. 25). Here again, dai contributes to the resulting pragmatic meaning due to the mirative value it can have when used in isolation, as discussed above. In examples (26) and (27)  As seen before, this pragma-dyad is prone to develop interjection-like values, and this holds also with regard to the second core value of ma dai: indeed, it can sometimes behave as a routinized exclamation of surprise which is graphically conceived as an independent unit, as in (28). The possibility of graphically (and phonetically) agglutinating the two components of the pragma-dyad is a revealing sign of the high degree of pragmaticalization this complex PM has undergone. To be precise, we found 634 occurrences of the univerbated form maddai in the corpus, frequently correlating with an interjection-like status, conveying either contrast or surprise. This means that formal coalescence ties in inherently with a routinization of pragmatic meaning, leading to the formation of an opaque interjection where the two components are not clearly distinguishable since the meaning of dai is at least partially bleached and the pragma-unit constitutes a semantically void slot. As a result, while the adversative semantics of ma contributes to signal counter-expectation, dai is reminiscent of its original lexical content pointing to the exchange of messages-as-object in the communicative exchange, and, more specifically, to an unexpected reception, this leading to the implied meaning of surprise ( Figure 6). Figure 6. Semantic components and pragmatic structure of the ma dai 2 pragma-dyad.

[[ma] CO-UNIT_CONTRAST + [dai] PRAGMA-UNIT_ RECEPTION ] COUNTER-EXPECTATION
Table 2, in turn, summarizes the distribution of the first 100 attestations of the ma dai pragma-dyad in the corpus across the different functions we have identified and described above. We found that the marker is very frequently employed as an interjection of contrast -a secondary development which, however, is already stable in use. The multi-functionality characterizing the ma dai pragma-dyad, with its two different core values, the development of interjections, and the frequent univerbation of the two components should all be interpreted as revealing signs of a high degree of pragmaticalization, productivity, and routinization, respectively. As we will see, the combination of these interacting factors -multi-functionality typically determines increase of use, and routinization is a by-product of token frequency -gives rise to the specific status of ma dai as one of the most productive instantiations of the ma 'but' + pm in the imperative pragma-dyad.

Ma piantala
In Italian, piantala is a relatively rude PM meaning 'stop it!, cut it out!, give me a break!'. It forms the 2 nd -person singular of the imperative of the verb piantare, literally meaning 'to plant; to abandon, to leave', plus an enclitic object pronoun la 'it' in the feminine. In our corpus we counted 1,224 attestations of piantala and 168 of ma piantala. Interestingly, within the set of data extracted from the corpus we found both cases in which ma piantala licenses different argument structures, thus showing a certain degree of syntactic integration, and instances where the marker clearly behaves as an interjection proper.
In the first case, ma piantala can be followed by an infinitive clauses introduced by di 'to', or can head a Prepositional Phrase featuring con 'with', probably under the paradigmatic influence of the synonymic expression smettila con ('stop, enough of'), which may have enhanced an analogic extension of an instrument-like Prepositional Phrase. We have found 32 tokens of ma piantala di + infinitive (ex. 29) and 13 attestations of ma piantala con 'with' (ex. 30). In both cases, the complex PM is used to order someone not to do something (e.g. criticizing, ex. 29; or continuing with a diet, ex. 30), i.e., to convey a FTA which overtly establishes a contrast with the interlocutor: (29) Ma piantala di criticare perché si può anche sbagliare 'but stop criticizing because anybody can make mistakes' (30) ma piantala con la dieta che poi sparisci del tutto!!!! 'but stop your diet or you'll disappear entirely!!!!' In the second case, ma piantala is syntactically independent and constitutes a graphically autonomous unit, usually indicated by exclamation marks, as in ex. (31) -which corresponds in the spoken language to a separate intonation unit in emotionally loaded exclamations. Alternatively, this PM can occur after an assertion with which the speaker rejects the validity of the interlocutor's claim. In (32), for instance, a previous interlocutor has probably accused the speaker of having voted for Berlusconi: the reaction on the part of the speaker is to deny it, also strengthening the illocutionary force of the denial with the addition of a suspended ma piantala as a harsh turn-closing device:

Ma va' and ma vieni
In PDI there is a constellation of deverbal markers that have pragmaticalized out of verbs of movement. It is often the case that verbs of movement develop pragmatic meanings in the world's languages, because they are inherently dynamic and naturally include a deictic anchoring in their semantics, since movement presupposes a change of location oriented towards a goal, which, in turn, fosters a contextual connection of the interlocutors with the shared communicative situation (cf. Radden 1996: 431). Heine and Kuteva (2002: 159-160) even mention a specific path of semantic change concerning verbs of going that can develop into what they call «hortative» imperative markers. This path is documented for Italian as well: the crucial role of the dynamic semantics of andare resulted, for example, in the pragmaticalization of the 2 nd -person singular imperative vai/va' 'go' as an emphatic marker that strengthens orders and requests. This process of functional enrichment basically rests on a metaphorical mapping licensed by the lexical semantics of andare: the literal meaning of proceeding across space has been metaphorically transposed onto the figurative one of proceeding to do something, since pushing somebody across space can be seen as driving somebody into action, with an exhortative sense (Fedriani / Ghezzi 2014). A clear example of the exhortative function performed by va' in directive acts is given in (33), cited in Fedriani / Ghezzi (2014), where the order leggi un po' questo e pensaci un po' is followed by an extra-clausal va', added as a sort of afterthought which strengthens the illocutionary force of the request.
Further pragmatic developments of verbs of movement include three interjections, namely evvai, ma va' (là) from andare 'to go' and ma vieni from venire 'to come'. Let us briefly look at the first, and then focus more extensively on the others, which all include ma in their dyadic combination.
Evvai constitutes the agglutination of the coordinative conjunction e 'and' plus vai, 2 nd -person singular imperative of andare, and a phonotactic doubling of [v]. Now, the coordinative semantics of e is crucial in shaping the global meaning of the resulting interjection: e, indeed, points toward an agreement, since the interlocutors are building the conversation in the same direction (you and me), which conveys the idea of parallel motion along the same argumentative lines. This is why evvai, start-ing from the general meaning of a coordinated parallel motion, came to be conventionalized as an expression of enthusiastic joy and appreciation, as shown in ex. (34) Frequently ma va' is reinforced with the addition of another deictic element, the locative adverb la' 'there', which further strengthens the metaphorical image of a centripetal movement directed away from the origo -i.e., stressing the idea of an interlocutor moving further away from the idea pursued by the speaker. The interlocutor's point of view if often reported in the form of a citation, and then refuted by the following ma va' là, which can also be added as a free-standing marker of disagreement with a holophrastic value, as in (37): (37) "Luxuria, partecipando e trionfando all'isola, ha spiegato a milioni di italiani che la realtà è diversa". Ma va là. '"Luxuria, by taking part in and by winning Isola [a reality show equivalent to "I'm a celebrity", CF & PM], explained to millions of Italians that reality is different". Do me a favour.' Here the contrastive meaning of ma is crucial in shaping the motion semantics of va' in the metaphorical sense of a divergent discursive move: "you go in an opposite direction from me", therefore, we have a disagreement. The semantic import of the constitutive co-unit and pragma-unit and the way in which they contribute to the co-construction of the resulting pragma-dyad is schematically represented in Figure 8. Interestingly, however, this is not the only pragmatic meaning developed by ma va'. When this interjection is uttered with a rising intonation, it carries a mirative sense, expressing counter-expectation and surprise. Once again, the adversative semantics of ma stresses the difference in perspective, as if the speaker were astonished to find out that the interlocutor was going in an unexpected direction, contrary to his previous assumptions. In such cases, ma va' can be translated as 'really?', often with a sarcastic nuance, as in (38), where the sardonic reading is emphasized by the ironic commentary non mi era sembrato 'that's not the way I saw it'.
(38) #frasedelgiorno "il tecnico non ha in pugno la squadra" Moratti. Ma va?! Non mi era sembrato '#claimoftheday "the coach doesn't have the team under his control" Moratti. Really?! I didn't realize it myself' Often, ma va' used as a sarcastic marker of counter-expectation is graphically inserted as a personal comment, in brackets, and its exclamatory quality is signalled by a long series of exclamation and/or question marks: (39) Il tipo mi risponde che io ho ragione (ma va???!!) ma che loro hanno ricevuto l'indicazione di procedere dall'ufficio anagrafe 'The guy answers me that I'm right (oh, really???!!), but that they received instructions to proceed from the registry office' Overall, we found 98 attestations of ma va' in the ItTenTen corpus. Their distribution across the two main core values of disagreement and counter-expectation is given in Table 3, which basically highlights that the two values are used with a very similar frequency. Let us now turn to the case of ma vieni, which constitutes a peculiar pragma-dyad because its pragma-unit, vieni, does not actually exist as an autonomous PM in PDI. Moreover, this pattern is the most recent and ephemeral PM to have entered this series, as we will see below. This case is of particular interest since it testifies to the productivity of the ma + pm in the imperative pragma-dyad, which shows that it is capable of attracting within its abstract template not only PMs already existing in the language, and presumably frequently activated in the mind of speakers, but also new items which do not carry a pragmatic meaning outside the constructional pattern of the pragma-dyad. But what is the functional core of ma vieni, and how did it emerge in Italian?
First, ma vieni is a much less frequent marker compared to the other cognate forms under consideration: in the corpus we only found 51 attestations. Its core function is that of expressing enthusiastic joy and satisfaction, linked to a sense of positive surprise. This positive connotation is clearly testified by the co-occurrence of an exclamations of happiness like evviva 'hooray' (ex. 40) and siiiiiiiii 'yesss' (ex. 41) and claims such as sono troppo felice and ho troppo goduto 'I am so happy' (ex. 41 and 42, respectively): (40) Evviva! Sono finiti i 6 mesi di prova. Ma vieni!!! 'Hooray! The 6-month trial period is over. Hooray!!!' Note that this highly routinized interjection is also used parenthetically (ex. 43), reflecting an independent intonation exclamation contour which is also clearly mirrored by the fact that this marker is often accompanied by graphic strategies of intensification, such as repeated exclamation marks and the use of capital letters (ex. 44-46): (43) ovviamente il sottoscritto è arrivato in finale (ma vieni) poi ho perso quando abbiamo fatto gli spareggi cmq decima posizione 'obviously the undersigned reached the final (hooray) then I lost in the playoffs at least I came tenth' As it is evident from the examples given above, ma vieni has a similar meaning to evvai, but it is more recent, as the indirect evidence given in (47) testifies, according to which 'nowadays you use more ma vieni' than e vai. In particular, thanks to the remark given in (48), the emergence of ma vieni can be dated precisely to the year 1981 and its emergence and spread linked to a TV catchphrase: (47) Una volta si diceva "...e vai!", oggi usa di più "... ma vieni!". Ma vai o vieni, è andata, ed è andata bene! 'They used to say "...e vai!", nowadays you use more "... ma vieni!". But go or come, it is gone and it went well!' It seems that social interaction, especially through the media, played a crucial role in the social embedding of ma vieni. This marker is indeed particularly frequent in sports contexts: in the corpus we scrutinized, we found that 8 occurrences out of 51 express joy and satisfaction for sporting achievements, as in ex. (49)(50) Ma vieni is also used in a very popular film comedy (Tre uomini e una gamba, 1997) that was very successful in Italy. The pragma-unit appeared in a specific scene which became a sort of catchphrase for young people in the 1990s. Probably the use of ma vieni by a famous actor had a catalyst function in triggering and reinforcing its spread and its productivity effects.
The question thus arising at this point is what is the possible source of this dyadic construction, since vieni does not exists as a PM or interjection on its own. In our view, ma vieni developed as an analogic form modelled on multiple sources which, crucially, co-exist as paradigmatic variants instantiating the same pragma-dyad, namely ma dai and ma va', all having, besides other possible values, a counter-expectation meaning. Therefore, they presumably played a substantial role in moulding the function of surpriseD joy to the new emergent form. As far as the specific connotation of joy is concerned, we believe that a possible motivation for its development and conventionalization is rooted in the lexical semantics of the verb venire and, more precisely, in its deicticity. While andare, which gave rise to va '/vai and ma va' (là), is basically oriented outwards, towards an endpoint which does not coincide with the speaker, venire is clearly centripetal, following Bourdin's (2003) distinction between motion oriented towards "otherness" (like "go" verbs) and motion towards "identity" (like "come" verbs). From this perspective, ma va' (là) accommodates the centrifugal frame <motion + otherness>, while venire can be better described as centripetal: <motion + identity>. Now, it is our contention that such different deictic orientations played a role in determining a pragmatic (although partial and subtle) divergence in the connotation implied by the various markers. While the centrifugal deixis entailed by andare triggered the development of pragmatic meanings pointing to a clear distance from the speaker (notably, disagreement), also along the lines of diverging expectations (neutral surprise), the centripetal orientation licensed by venire tended rather to encourage a positive reception, and on the part of the speaker a surprise which came close to the speaker's subjectivity, thus entering the personal sphere of the origo. The semantic-pragmatic structure of ma vieni is summarized in Figure 10. Summing up, although vieni does not exist as pragmatic resource on its own in PDI, due to its lexical meaning pointing to a centripetal deictic orientation it has been attracted to the pragma-dyad, as a novel pragma-unit compatible with it on account of the compelling paradigmatic analogy exerted by functionally neighbouring constructions. Since these constructions, acting as multiple source models, instantiate the same abstract template, they shared a certain degree of internal pragmatic coherence which presumably turned out to be a key factor in enabling the pragma-dyad to attract new pragma-units. This is, in sum, a clear sign of the schema's productivity.

Paradigmaticization and productivity of the ma + PM in the iMPerative pragma-dyad
In the functional description given in Section 3 we showed how the ma + pm in the imperative pragma-dyad emerged and conventionalized as a kind of pragmatic construction, which, as Fried / Östman (2005Fried / Östman ( : 1773 put it, «specifies not just morphosyntactic or lexical-semantic information, but also conventionalized pragmatic and interactional features». In the case at issue in this paper, such a construction is made up of two recurrent constitutive parts which, as we have seen in detail, functionally shape the resulting instantiations of the general schema. Now, it is worth stressing here that such instantiations can be seen as cognate outcomes of a coherent paradigmatic set, «which are (unconsciously) perceived by the language users to be closely related to each other in the constructional network» (Traugott / Trousdale 2013:14).
Such a process of paradigmaticization further promoted semantic and functional convergence, which probably produced, in the words of Octavio de Toledo (2018), «a set where the evolution of any individual member largely depends on the evolution of the others», something «which appears to suggest synergic effects attributable to the group's internal dynamics». In his interesting study on a paradigm of Discourse Markers built around the Spanish marker bien, Octavio de Toledo (2018) has indeed shown that the different instantiations acted as mutually «supporting constructions» (De Smet / Fischer 2017) that guided the development of each individual marker in the direction of the core semantic and functional features most compatible with the rest of the paradigm. Along similar lines, the frequent combination of ma + different pms in the imperative as possible fillers led to an increase in token frequency of the pragma-dyad, since token frequency determines the degree of entrenchment of a single word or construction (Croft / Cruse 2004), which therefore becomes able to sanction new extensions -and this in turn determines an increase in type frequency, as the case of ma vieni has neatly shown. Needless to say, an increase in type frequency further strengthens the mental representation of the construction (see Hoffmann / Trousdale 2011: 15).
Building on these observations, the emergence of a structured paradigm and the productivity of the pragma-dyad can be summarized as follows. In view of its semantic and structural consistency the pragma-dyad progressively gave rise to different types (ma scusa, ma dai, ma piantala, ma va'), eventually attracting a pragma-unit which does not exist in isolation as a proper PM in PDI (ma vieni). This testifies to two intertwined aspects, namely the pragma-dyad's productivity, in terms of both structural extensibility to new instances entering the paradigm, and analogy, which, according to Barðdal (2008), is a side-effect of the pragma-dyad's high internal coherence, a crucial feature that makes it possible to target synonymous or semantically close verbs which may be compatible with the general procedural schema, such as vieni (a motion verb like va', which however differs in deictic orientation). Moreover, the pragma-dyad was able to expand to include to new pragma-units, which formally differ from the core ones, since they fill the slot with heterogeneous elements, such as politeness markers (ma per favore 'but please', see Fedriani 2019), verbs in the 1 st -person plural subjunctive (ma finiamola 'but let's stop it' and ma andiamo 'but let's go'), and interjections (ma basta 'but stop'), which were in some way 'coerced' by the meaning conveyed by the pragma-dyad and became semantically compatible with it. In our view, there is a strong association between ma scusa and ma per favore, two complex markers both pertaining to the functional domain of politeness; between ma piantala and its synonymic variants ma finiamola and ma basta; and between ma vieni and the equifunctional marker ma andiamo, which expresses the very same meaning of joy and appreciation corresponding to English 'hooray' and typically found as an exclamatory interjection only used in sports contexts -note that indirect evidence for the synonymy between ma vieni and ma andiamo comes from the fact that they are associated with the same meme image on the web (Figure 11).
Crucially, the links holding between ma vieni and ma andiamo on the one hand, and between ma piantala, ma finiamola and ma basta, on the other, can be better understood as cases of semantically motivated extensions of a recurrent instantiation to synonymic elements. This can be regarded as a paradigmatic analogical process which, following Barðdal (2008: 89), is conceived as «based on lowest possible type frequency, i.e. one, and highest degree of semantic coherence, i.e. full synonymy between the source and the target item». Drawing on these premises, the inner structure and the productivity of the pragma-dyad can be depicted as shown in Figure 12, where bold lines denote productive instantiations built on already existing PMs; dashed lines indicate outcomes modelled on verbs which do not have a pragmaticalized counterpart, but were attracted to the pragma-dyad through paradigmatic analogy; and dotted lines represent indirectly productive relations, in other words, extensions of the schema to heterogeneous pragma-units. As the pragma-dyad was extended to new items, the result was probably an increase in type frequency, which determined a parallel growth in token frequency, and, ultimately, the tightening of the pragma-dyad as a coherent network built around the procedural meaning of interactional contrast.
In conclusion, the different forms in which the pragma-dyad investigated here can manifest itself can be better seen as alternatives «available on the axis of similarity and choice» (Traugott / Trousdale 2013: 197), belonging to a structured network and undergoing a productive process of paradigmaticization, one of the multiple domains where linguistic change -including pragmatic change -can take place.

Conclusions
In this paper we have shown that the notion of pragma-dyad has the advantage of accounting for the paradigmaticization of a set of formally and functionally cognate pragmatic elements, which all result from the recurrent association of two autonomous items with two different meanings M 1 and M 2 , both existing as independent units in a given language but that, once combined together, give rise to a more abstract schema with a new meaning M 3 . Such an M 3 meaning is built on, but different from, M 1 and M 2 -namely, a presupposition of contrast between two opposing views. Once the pragma-dyad's meaning M s became entrenched enough in the mind of speakers, it eventually attracted new items which cannot be used in isolation, such as vieni. The case of ma vieni is noteworthy because it shows that the pragma-dyad exists at a high level of abstraction, given that its pattern can be replicated through paradigmatic analogy and whose pragmatic sense is licensed precisely by the pragma-dyadic meaning. In addition, what is more distinctive about the present approach is that we believe that the notion of pragma-dyad can be fruitfully applied to a variety of similar cases of pragmatic paradigmaticization. To mention a case in point from Italian, in our view the pragma-dyadic approach creates a window of opportunity for a systematic account of a cognate schema, featuring the conjunction e 'and' plus a set of pragmatic fillers, some of which have been investigated in this paper (e.g., the above mentioned case of e + vai: evvai, ex. 28; dai: eddai, among others).
Moreover, the pragma-dyad notion helps us account for the structured polysemy developed by the co-unit ma within the pragma-dyadic network. As we have seen, ma can thus express different nuances actualizing the core function of interactional contrast, terms of mock politeness, disagreement, counter-expectation, depending on the different fillers it can combine with. Thus, the co-unit and the pragma-unit dynamically interact in the co-construction of the resulting complex meaning, since they are actively involved in a mutual process of semantic and pragmatic enrichment. Although as they are part of the same coherent network, we have seen that each complex marker has its own story in this respect, as summarized in Table 4. Lastly, we have seen that the pragma-dyad can give rise to different outcomes, placed along a notional continuum of pragmaticalization, ranging from PMs to interjections: compare, for instance, the case of ma scusa, which admits different orders (scusa ma) and also retains some inflectional properties (ma scusate, 2 nd person plural), with that of ma piantala, and ma vieni, which are frozen, partially bleached, and express emotive reactions (irritation, surprise, joy), functioning as holophrastic exclamations, which in some cases can even be univerbated (maddai). Ma va' and ma dai, in turn, exhibit an ambivalent status, and demonstrate how these complex markers constitute a fluid domain, subject to discourse variation and open to novel coinages and ephemeral changes, which may constitute temporary forms linked to cultural products and passing fashions. This is particularly evident in the case of ma vieni, where we have seen how the media can represent relevant sources for the birth (and decay?) of form-function configurations.