Romanian de fapt – from adjectival adjunct to attention marker

This research traces back the development of the Romanian phrase de fapt (‘in fact, actually, indeed’), based on written and oral corpora. De fapt has been attested in Romanian since late 19 century; chronologically, it is the last of the three Romanian adverbial expressions (alongside în faptă and în fapt) that went through all the stages of the grammaticalization cline proposed by Elizabeth Traugott for this type of adverbs. However, we consider that this phrase actually goes even further by becoming, in press headlines, an attention marker (Fraser 2009: 297), thus joining the category of să vezi ce s-a întâmplat (‘you won’t believe what has happened’). Thus, in press titles such as Cu ce femeie a petrecut aseară Pepe, de fapt (‘Who is the woman Pepe actually spent the evening with’), de fapt loses its contrastive discourse marker rhetorical function of contrasting with a previous element and acquires a new function, i.e. of inviting the reader to read a story that (s)he would have otherwise overlooked. In this type of occurrences, de fapt acquires, for the first time, an intersubjective value.

The papers that diachronically discuss the Fr. de fait, en fait or the Engl. in fact highlight a three-stage evolution, in which the adverbial phrase moves from being an internal adverb to being a sentential adverb and, finally, a DM, with a wider scope aiming at a three-element sequence: <S1 -DM + S2>. This study follows in the footsteps of previous similar analyses and aims to determine to what extent we could talk about a fourth stage in the evolution of the Romanian form de fapt. The next section reviews the stages of the cline proposed by Traugott, which is also an opportunity to discuss some terminological concepts; the third section is a diachronic analysis of the series în faptă/în fapt/de fapt; the fourth section lists the values of de fapt in present-day Romanian, which allows us to deal, in the fifth section, with its particularities in press headlines. The sixth part is dedicated to the conclusions.

The theoretical framework
This research uses the term discourse marker within the meaning proposed in the works of Bruce Fraser. For this author, DMs are a class of pragmatic markers, the latter being identified as linguistic signals related to "non-propositional part of sentence meaning" and "linguistically encoded clues which signal the speaker's potential communicative intentions" (Fraser 1996: 323). According to Fraser, there are four classes of pragmatic markers: basic pragmatic markers (I promise, please), commentary pragmatic markers (fortunately, frankly), discourse markers (but, and, so) and discourse structure markers (in summary; returning to my previous topic; look, now) (Fraser 2009: 295-297).
For Bruce Fraser, a specific element is a DM if it can be accepted in the sequence <S1 -DM + S2>, where S1 and S2 are discourse segments and illocutionary acts. In order to function as a DM, the respective element must be a lexical expression, which is part of S2 but has no contribution to its meaning, and it must only point out to a specific semantic relationship between S1 and S2 (Fraser 2009: 297-299). The first condition excludes syntactic structures, prosodic features and non-verbal expressions from the category of so far. The research splits at this point: this chapter focuses on a theoretical discussion of these stages as an opportunity to go through the terminology and the defining features of each stage, while chapter 3 presents the diachronic perspective, with examples from different corpora.
The Romanian DMs subsumed to the DE FAPT 2 type (în faptă, în fapt, de faptă, de fapt, cu faptul) are loan translations of the French corresponding expressions (dans le fait, en fait, de fait, par le fait, en effet, le fait est que). The nouns used in Romanian phrases are fapt (< Lat. factum) and faptă (< Lat. facta) 3 . Thus, fapt 'fact' and faptă 'deed' represent the starting point in the development of the structure we investigate. STAGE 0 faptă0/fapt0nouns, syntactic constituents of some prepositional phrases, freely occurring with different prepositions. In this type of contexts, fapt/faptă display the following syntactic features: it can get determinersa definite article as an enclitic determiner and inflectional marker (1), an (implicit) demonstrative determiner (2) or it could be interpreted as a plural (3-4) 4 : (2) Ca să ne dăm seama that SĂ.SBJV CL.REFL.1PL assume.SBJV.1PL representation de fapt să aplicăm sistema la particular about fact SĂ.SBJV apply.SBJV.1PL method to individual To easily understand this fact, let's turn from general to individual. (Eminescu,P,XIII,409) 2 DE FAPT stands here for the whole series of Romanian expressions with the same meaning and similar evolution: în faptă, în fapt, de fapt. The other two, cu faptul and de faptă, stopped at the first stage of development. 3 Lat. facta is the plural of Lat. factum. In the oldest Romanian texts faptă does not display plural morphological features. It only occurs as a feminine singular noun (DA, under faptă). 4 I have introduced interlinear glosses (following the Leipzig interlinear glossing rules, available at https://www.eva.mpg.de/lingua/pdf/Glossing-Rules.pdf) for all shorter examples, where the sequences în faptă, în fapt and de fapt go through the first 3 stages of their evolution. I have only partially glossed the longer examples (where the glossed segment is highlighted in italics, and the gloss appears in brackets). However, I have left the very long examples unglossed, because glossing morpheme-by-morpheme is almost irrelevant for a study in the field of pragmatics. For example, in this study, what is important is the relation in terms of meaning that the discourse marker de fapt creates between the discourse segment that precedes it and the discourse segment that follows it, and not the morpheme analysis of the text itself. 5 Our italics throughout the text, except where we indicate otherwise.
( Even when the other element of the pair is not lexicalized, it can be contextually inferred, as in (6), where the sequence legea ... în faptă suggests de jure/de facto pair, or in (7), 6 "Modern usages of au fait, de fait and en fait (Engl. in fact) are the result of a threefold process: the lexicalization of a prepositional phrase into a single functional unit entailing the grammaticalization of the three forms as adverbs, followed by a process of pragmaticalization of the adverbs as pragmatic markers." (D'Hondt 2014: 236) where the succession sfatul ... în fapt refers to the pair în vorbă/în fapt (Engl. in speech/in fact).

STAGE 2
DE FAPT2at this stage, the second grammaticalization occurs: the scope of DE FAPT2 is now the whole sentence, it functions as a sentence modifier. Schwenter & Traugott 2000 discuss stage 2 of the Engl. in fact and notice that, as a member of the epistemic certainty expressions, it "indexes the speaker's strong commitment to the proposition, and is higher on a scale than probably and possibly, in the same way as strong epistemic must is higher on the epistemic scale than may and might" (Schwenter & Traugott 2000: 12). A similar remark belongs to Taglicht 2001 referring to actually2: the speaker who uses it in a certain utterance "is envisaging a scale of properties with the content of that phrase at the top, and implying that any property below it on the scale would make the expression too weak" (Taglicht 2001: 2). For Traugott 1999, the responsibility of phrasing the utterance preceding DE FAPT2 belongs to the speaker. Thus, the speaker signals a contrast either a) with the previous utterance, which refers to an objective reality, and to the expectations of the listener/reader related to it, or b) with the expectations that the speaker creates. The speaker's commitment to the content of the sentence does not change, but the way in which this content is expressed significantly differs in the two situations above.
The following examples underline the first case: what is obvious here is the speaker's/narrator's commitment to the content of the utterance including în fapt2 or în faptă2. It is also obvious that the reader's expectations after the first sentence/verse are contradicted by the speaker/narrator in the second sentence/verse: Because he was kind of a reckless, madcap guy But his deeds don't prove him to have been a trickster. (Pann,NH,204) The difference between the use of these adverbs as sentence modifiers and their use as stage-1 adverbs, in which the scope of the adverb is restricted to the syntactic group that it belongs to, can be clearly seen when used simultaneously in the sentence: "but, în fapt2 (i.e. 'in reality') the house of amortization failed to fulfil its în fapt1 purpose (i.e. 'the real/initial purpose')"; "he doesn't prove în faptă2 ('in reality') to have been a în faptă1 trickster (i.e. 'a real/true-born trickster')". The examples above have been intendedly phrased in this particular way, but they have not been highlighted, and their degree of subjectivity is more vague than in (10) below.
In the second group of examples, the speaker is more subjectively involved. In this case, (s)he "sets up false scenarios and then shows that the assumptions manifest in them are wrong, inappropriate to the occasion, etc." (Traugott 1999: 183). The evolution towards stage 2 implies a subjectification process, because "the development of the EA [epistemic adverb] meanings of indeed, in fact, and actually is in each case an example of subjectification" (Traugott & Dasher 2002: 174). The false scenario that Traugott talks about involves a counter-expectation, which is a "rhetorical strategy", as noticed by the quoted author (Traugott 1999: 178). It "is a matter of point of view and is therefore an example of the subjectivity of language" (Traugott 1999: 179). Example (10) is a typical case of constructing a counter-expectation: (10) Atunci, dacă un quadrilater nu este nici quadrat, nici "dreptunghiu" (sau mai bine oblong), nici "paralelogram" (sau mai bine romboid), nici romb, ar trebui să nu poată fi alta decât trapez. În fapt însă mai poate fi alta (in fact.ADV however also can be.INF another) adecă o figură planimetrică închisă în patru laturi fără nici un paralelism [Maiorescu's italics]; un asemenea quadrilater se numește trapezoid […]. Then, if a quadrilateral is neither a square, nor a 'rectangle' (or rather oblong), nor a 'parallelogram' (or better a rhomboid), nor a rhombus, it can only be a trapezium. But, in fact, it could also be something else, i.e. a four-sided planimetric shape without any parallelism; such a quadrilateral is called a trapezoid […]. (Maiorescu,L,33) The epistemic value of în fapt is obvious in (10): în fapt, însă, mai este una (= 'for sure, there is another one'). In his example, the author conveys the contrast differently: he lists all types of quadrilaterals and he eliminates, one by one, those whose opposite sides are parallel (square, rectangle, parallelogram, rhombus), retaining (apparently after having exhausted all possibilities) the quadrilateral with only two parallel opposite sides, i.e. the trapezium. This staging is intended to highlight (as the emphasis in italics shows) the fact that we would be wrong to judge things in this way, because we might miss a possibility that we would not have thought of (but which exists objectively), i.e. a quadrilateral in which there is no parallelism between the pairs of sides that make it up. Those to whom this lesson is addressed may know very well, from geometry classes, that such a figure exists but the author assumes that they do not. Usages based on counter-expectation have a certain degree of subjectivity, lower in (8) and (9) and higher in (10). As pragmaticalization is in close relation with subjectivisation, we can talk here about the beginning of the process of pragmaticalization.
STAGE 3 DE FAPT3contrastive DM. The scope of this type of DM is a more complex structure than the clausal adverbs, namely <S1 -DM + S2>. The process of pragmaticalization started at the previous stage continues: the subjective dimension of de fapt3 is stronger, emphasizing the speaker's perspective over the events 7 . Stage 3 implies an even greater 7 Pragmaticalization is a process during which a certain unit undergoes formal (phonological, morphological, syntactic) and semantic changes (see, for example, Dostie 2004: 35-37, for a review of the principles of pragmaticalization). However, our study focuses mainly degree of subjectification, because the DM signals that the upcoming utterance is a stronger argument (on the rhetorical scale, Schwenter and Traugott 2000: 12) than the one preceding it. While DE FAPT2 "combines aspects of the meaning of strongly epistemic adverbs like certainly and adversative adverbs like however", the lexical field to which DE FAPT3 belongs "consists of such discourse markers as what's more, indeed." (Schwenter and Traugott 2000: 11, 12).
DMs are frequently used in polemical communicative situations, where the speaker replies to an interlocutor's speech, which the former alludes to or recalls in his own speech:  (Kogălniceanu,CU,5). This disregard of the teaching activity has as a direct result the fact that the students no longer attend their classes. Example (11) can be rewritten as follows: "I thank the Minister for his report on the activity of the University of Iași. In fact, the University of Iași no longer exists". The situation that the speaker refers to does not confirm the historical reality (since its establishment in 1860, the University of Iași has never ceased to exist, neither de facto nor de jure), but belongs to the rhetorical and subjective order of the discourse (confirmed once again my conviction is a subjectification discourse constituent).

STAGE 4
DE FAPT4 -If de fapt3 is a DM and differs from de fapt2, which is not yet a DM, it also differs from de fapt4, which is no longer a DM. De fapt4 belongs to another class of pragmatic markers, illustrating a step forward in terms of rhetorical use, and it displays a higher degree of (inter)subjectivity compared to the previous stage (i.e. it acquires an intersubjective value). We shall discuss the features of this type of pragmatic marker in section 5.
on the ever deeper subjectification process that takes place during the transition from de fapt2 to de fapt3, and then on intersubjectification, during the transition to de fapt4. For D'Hondt 2014: 259-260, the fact that en fait actually acquires a pragmatic role means that "the introduction of en fait into the discourse creates a subjective dimension, emphasizing the speaker's perspective." Traugott 2010: 38 also talks about a strong correlation between grammaticalization (which, for the quoted author, also means pragmaticalization) and subjectification.

Diachronic overview
3.1. The lexicographical recordings of the elements of the Romanian DE FAPT series, în faptă/în fapt/cu faptul/de faptă/de fapt begin in ECR (1931). It only mentions de fapt, which receives three synonyms ("în realitate, aievea, efectiv", Engl. 'in reality, as a matter of fact, in practice'); however, it does not specify its morphological nature, it does not provide examples of usages and it does not offer a different etymology from the entry word (fapt < Lat. factum). DA (1934) recordsunder faptcu faptul and de fapt ("phrase", without any further specifications). It says nothing about the third member of the series, în fapt, although faptă also includes the "expression" în faptă, "recently replaced by de fapt", which would have theoretically required the indication of în fapt under fapt. This is even more bizarre, as in Maiorescu's Logic, quoted by DA for de fapt (de fapt1, with a single occurrence in Maiorescu's entire work), în fapt1 has 6 occurrences, and în fapt2, 15 occurrences. In contrast, DA mentions cu faptul, for which we found only one occurrence in all the works in the corpus, namely in Ghica's Scrisori către V. Alecsandri. The Fr. etymology proposed by DA for de fapt is dans le fait, en fait, par le fait, en effet and réellement; however, it does not consider the Lat. de facto, member of the de jure/de facto opposition, with legal usage, as an option, which could have occurred in the evolution of de fapt at least from a certain moment onward. Scriban (1939) also mentions two members of the series: în fapt and de fapt; the latter is recorded only in opposition to de drept, opposition illustrated as follows: Ludovic XIII era rege de drept Louis XIII be.IND.IPF.3SG king of law.ADJ dar de fapt era Richelieu but of fact.ADJ be.IND.IPF.3SG Richelieu Louis XIII was king as of right, but the real king was Richelieu which would be equivalent to a connection, merely suggested, between the pairs de drept/de fapt and de jure/de facto). In addition, the pair rege de drept/(rege) de fapt clearly indicates that Scriban focuses only on the stage de fapt1 and only on its adjectival usage. None of the three dictionaries provides examples of de fapt/în fapt/în faptă used as adverbial phrases (cu faptul is only adverbial).
3.2. Therefore, the dictionaries provide incomplete and confusing information and do not even give a vague idea about the evolution of the forms discussed herein. We try to formulate some observations regarding the evolution of the Romanian forms, starting from the data provided by a corpus that covers (considering the years of their writing) the 1829-1947 period. The corpus we chose for the early 19 th century includes historical writings -the capital work of N. Bălcescu, prosethe complete work of Negruzzi, and poetrythe writings of Alecu Donici and Povestea vorbii and Năzdrăvăniile lui Nastratin Hogea by Anton Pann, which total almost 2,200 pages. For the late 19 th century, we chose several volumes of parliamentary speeches -V. Boerescu, Ion C. Brătianu, Titu Maiorescu, memories -Ion Ghica's Scrisorile către V. Alecsandri, a travel diary -I. Codru-Drăgușanu, Maiorescu's Logic and all of Eminescu's newspaper editorials, i.e. a total of 9,000 pages. Finally, for the early 20 th century we selected for our corpus: parliamentary speeches -Vintilă Brătianu, Armand Călinescu, Nicolae Iorga, a university course by G. Călinescu and a part of his editorials, i.e. 2,400 pages, which means a general total of 13,600 pages. The data provided by the corpus allow us to make observations related to the history of the forms subject to discussion (3.2.1.), their combinatorial availability (3.2.2.), their relation to stage 0 (3.2.3.), as well as observations about their adjectival and/or adverbial use (3.2.4.) and their distribution according to stylistic criteria (3.2.5.). A brief comparison between the evolution of the Romanian forms and that of their French counterparts (3.2.6.) is also relevant for the present discussion.
3.2.1. The oldest Romanian form is în faptă, used in the early 19 th century with adverbial value and very rarely as a DM; it is competed by în fapt, widely used in the second half of the 19 th century, with multiple occurrences for the stages în fapt1 (mainly adverbial), în fapt2, but less frequent for în fapt3. De fapt developed in parallel in the 19 th century and it was mainly used as an adjective, rarely as an adverb and very rarely as a DM. The use of în fapt and de fapt radically differs in the first five decades of the 20 th century, when în fapt loses ground in favour of de fapt.
The examples prove that în faptă was used between 1800 and 1880. In our corpus, the first example in which în faptă appears as a DM dates to around 1840, in Alecu Donici's fables (example (12)). This suggests that Romanian followed a particular path of development, as the pragmaticalization of în faptă could be influenced neither by en fait (the first example of en fait3 in D'Hondt's corpus is dated 1855) nor by de fait, because "so far, de fait does not seem to have developed really pragmatic uses." (D'Hondt 2014: 260): To praise themselves in vain Some say: 'we've been working hard'. But actually they are working, As if they were a ploughing fly 8 . (Donici,S,78) The form în fapt, frequently used between 1850 and 1900 9 , becomes a DM (în fapt3) around 1880. In our corpus, it is attested for the first time in Eminescu's editorial, in 1878: (13) Lăsînd doparte acuzările de rea-credință, vorba ceea: "el dă, el țipă", prin care Românul voiește să inducă de mai înainte pe public în eroare, prevenim numai pe confrații de la Românul de a nu ne sili să ne întoarcem la argumenta ad baculum cu cari am fost adesea nevoiți a ne servi față de acești iubitori de adevăr confrați. În fapt însă, daca renunțăm de a întrebuința o manieră mai viguroasă de a discuta, n-o facem doar pentru că Românul ar fi încetat de a merita un tratament părintesc, 8 Donici's fable corresponds to Aesop's fable The Fly on the Wheel.

9
În fapt has never ceased to be used in Romanian, as evidenced by the examples in contemporary Romanian, where în fapt1 and în fapt2 continue to be used especially in the legal language, and în fapt3 faintly competes de fapt3; during the period we refer to, în fapt in general and în fapt3 in particular recorded the highest frequency rate. ci pentru că ne-am săturat noi înșine de a mai îndrepta un lucru strîmb prin chiar firea lui. Leaving aside the accusations of dishonesty according to the saying: 'he who strikes shouts the loudest', by which Românul wants to mislead the public, and we only want to prevent our fellows from Românul not force us to return to the ad baculum argument which we have often had to address to our truth-lover fellows.
In fact, if we give up using a more vigorous way of conducting an argument, we do so not only because Românul would have ceased to deserve parental treatment but because we are tired of straightening by ourselves something crooked by its very nature. (Eminescu,P,X,109) 10 Finally, de fapt, used from 1850 up to this day, becomes a DM (de fapt3) around 1880. In our corpus, it is first attested in Kogălniceanu's work, in 1877, example (11). The moment în fapt and de fapt reach stage 3 is subsequent to the one when the Engl. in fact becomes a DM (around 1815, Traugott & Dasher 2002: 171, figure 4.2.), but it is comparable to the one on which the Fr. en fait acquires a pragmatic value ("in our corpus, en fait appears with an essentially pragmatic value starting with the second half of the 19 th century", D'Hondt 2014: 258). This is the time for a necessary clarification regarding the interpretation of the forms that we examine. From a historical perspective, an example like (14)  A watch dog asks Juju, a pet dog, the reason behind the difference in treatment that he gets from his master. This question is likely to probe different levels of development of în faptă in the quoted excerpt. The present-day reader can interpret ce slujbă împlinești în faptă (lit. 'what services do you indeed carry?') as a redundant periphrasis, or an emphasis on the difference between appearance and essence ('is there really a reason for such discrimination?') or a contradiction between the speaker's utterance and reality ('actually, I wonder if there are any other hidden reasons except the ones you mention'). Such examples, which allow at least two readings and are quite frequent in the writings of that time, highlight the difficulties in interpreting such fragments (and the danger of overinterpretation), but seem to sometimes capture in statu nascendi a stage-1-to-2 or stage-2-to-3 semantic change of the structures that we discuss.

3.2.3.
Many other examples prove the connection between stage 0 and stage 1 of these forms. This relationship stands out if we go through the data in Table 1: The figures indicate a direct correlation between the frequency of the nouns faptă0, fapt0 and the frequency of the phrases în faptă1 and în fapt1. In Bălcescu, Pann and Donici, fapt has zero occurrences, and in Negruzzi and in the translations made together with Donici, this noun appears only in an idiomfaptul zilei (lit., 'the beginning of the day, dawn'). This is reflected in the absence of în fapt1 from the writings of the authors that we selected for the first half of the 19 th century.

3.2.4.
In Romanian, the stage 1 forms are likely to be both adjectival and adverbial adjuncts. There is a clear distinction between the two types of adjuncts in terms of frequency, for each form independently, as well as compared to each other. In the first half of the 19 th century, în faptă1 behaves exclusively as an adverbial adjunct. In the second part of the century, we identified 10 occurrences as an adverbial adjunct and only two as an adjectival adjunct. În fapt1 is used similarly, with 59 occurrences as an adverbial adjunct and with only 5 as an adjectival adjunct. Contrastingly, during the same period, de fapt1 occurs mostly as an adjectival adjunct (75 occurrences) and less as an adverbial adjunct (only 13 occurrences). These figures indicate a real competition only between the adverbial expressions în faptă1 and în fapt1, while în fapt1 and de fapt1 seem to be used preferentially by the authors either as an adverbîn fapt1 (59 occurrences, compared to only 5 as an adjective), or as an adjectivede fapt1 (75 occurrences, compared to only 13 as an adverb), as summarised in Table 2: The frequency of în fapt2 and de fapt2 in the late 19 th century is directly proportional to the frequency of adverbial în fapt1 and de fapt1; in Maiorescu, for example, the 39 occurrences of în fapt2 must be directly related to the large number of occurrences (18) of adverbial în fapt1; similarly, the almost complete absence of de fapt2 must be correlated with the small number of adverbial occurrences of de fapt1 (2 occurrences in 2,873 pages of Maiorescu's work). In the writings from the early 20 th century, we witness a paradigm shift: în fapt2 and în faptă2 are actually abandoned in favor of de fapt2counting 46 occurrences. These must be directly connected with the more frequent use of adverbial de fapt1 (11 occurrences), which represents the prerequisite for the appearance of de fapt3 (20 occurrences). In the early 20 th century, following the change of generations, de fapt takes over all the tasks of în fapt, as shown in Table 3:  (Maiorescu,DP,IV,535) The scope (in terms of discourse genres) of în fapt2 is therefore wider than that of the corresponding DM în fapt3.
3.2.6. The study of the data provided by the corpus suggests that Romanian followed its own path in the historical development of these forms, compared to French, which influenced Romanian only at the beginning of the 19 th century. French developed three concurrent and congeneric forms, au fait, de fait and en fait, attested in the 14 th century (au fait and en fait) and in the 15 th century (de fait), with a common origin (Lat. factum) and similar development, but with three different values in contemporary French. Thus, "au fait is essentially a discourse organizer […]. The sentence introduced by au fait expresses what is missing in the previous sentence, considered crucial for the continuation of the speech" (D'Hondt 2014: 242). It is preferred in interrogative phrases and interpellations, which imply an intersubjective value. On the other hand, de fait cannot introduce a new subject, it reformulates a previous point of view and refers to a previously established fact, thus looking backward; it can be seen in non-contrastive contexts and it introduces a confirmation or a clarification. Finally, en fait has the most extensive usages of the three forms and is the most frequent one; as a sentence adverb, "en fait has an epistemic value close to en réalité and it signals a particular contrast with the immediately preceding context or with the expectations created" (D'Hondt 2014: 247). At the textual level, as a connector, "en fait introduces a contrast in the broad sense compared to the previous sentence, i.e. a construction, a specification, a redefinition or a correction of the previous sentence" (D'Hondt 2014: 248). As a contrastive DM, en fait refers to new facts, it introduces a break with regard to the prior discourse. Of the three forms, en fait went the farthest on the path of its development as a pragmatic marker, and "the pragmaticalization of en fait continues even today" (D'Hondt 2014: 260). Therefore, French displays a pattern with three congeneric forms and parallel evolution. At present, these three share in a particular way the field of pragmatic markers, each one displaying a specific usage. The clear distinction between two of the meanings of the etymon fait: fait2 'fact, what happened, what exists' (the content of the forms de fait and en fait) and fait3 'the thing, the subject, the matter that it is all about' (the content of the form au fait) as well as the primary meaning of the preposition that each of these includes: à (destination orientation), de (distance from the point of origin), en (the place in itself) equally contributed to the development of three distinct types of usage for the French phrases (D'Hondt 2014: 238-239). Contrastingly, Romanian made no distinction between the semantics of fapt, and neither did it interpret differently the two nouns fapt and faptă (as it can be easily inferred from the similar and, for a short time, parallel evolution of în faptă and în fapt). Furthermore, unlike French, where the syntactic pattern involves a definite article (au, from au fait, comes from à le), Romanian did not differentiate between the syntactic patterns with a definite article (cu faptul) and those lacking a definite article (specified by different prepositions, în and deîn fapt, de fapt) and interpreted them alike.
Therefore, French encouraged the diversity of the forms and their preferential use, while Romanian exploited the polysemy of a single form, that changed from one epoch to another: an expression evolved, fully reached (or not) its potential, then it was abandoned in favour of another formthat had been its competitor for a whilewhich, in its turn, displayed the same development pattern.
An interesting distinction between the usages of en fait is made by Wandel 2017, who demonstrates that this framing adverbial has two kinds of usages in French, as "its framing scope (Fr. portée cadrative) is related either to the propositional content of the utterance or to the enunciation itself" (Wandel 2017: 5). Thus, if there are contextual indicators in S1 that signal a different perspective, at content level, from that existing in S2, then the framing scope of en fait is propositional. The set of contextual indicators is heterogeneous; it can include proper nouns indicating another speaker (Mr. Bertrand), evidentials (according to X), verbs of opinion (to consider, to believe, to think), adverbs referring to theory or appearance (in principle, theoretically, apparently), verbal modes or tenses expressing uncertainty (such as conditional or imperfect). In contrast, the framing scope of the discussed marker is enunciative if en fait "marks a change of perspective at the enunciative level, indicating the speaker's positioning with respect to [previous] utterances within the discourse" (Wandel 2017: 20) without such indicators in S1. In the latter case, en fait's ability to set a contrast weakens, and it may capture other valuesfor example, one of reformulation marker. The cited author believes that there is a correlation between the propositional/enunciative distinction and different discourse genres, in the sense that "more 'informative' genres, which do not explicitly evaluate opinions or perspectives (encyclopaedias, scientific texts, etc.), contain mostly propositional forms for structuring information and fewer enunciative forms, while the more 'argumentative', rather oral genres (political speeches, oral conversations, online verbal exchanges, etc.) need both enunciative forms to indicate the relevance of an argument and propositional forms to structure blocks of information" (Wandel 2017: 26). The evolution of Romanian forms shows that contextual indicators such as those mentioned above in 3.2.2. are characteristic for stage 1 of în faptă, în fapt and de fapt. At later stages of their evolution, the specific contrast for these discourse markers tends to be achieved through other types of indicators: rhetorical strategies such as the one illustrated in (10) or the achievement of a contrast between the utterances of a previous speaker, rendered in Direct discourse, as in (11), Indirect discourse, as in (25), or Narratted discourse, as in (12) and (26), and the statements of the actual speaker, contained in S2. Finally, enunciative usages of de fapt lose their contrastive value, as in (28)-(30). The three stages of de fapt (prepositional phrase, sentence modifier, discourse marker) are parallel and synchronous forms in current Romanian, but a study of their use in relation to the informative/argumentative distinction has not yet been made. As we have briefly shown in 3.2.5, it seems that, in Romanian, în fapt2 and în fapt3 can also be differentiated according to the discourse genres in which they appear, în fapt2 being specific to informative genres, and în fapt3 to argumentative ones. However, this is only a hypothesis and it needs to be tested for the usages of de fapt in contemporary Romanian.

The current status of the de fapt DM
The Romanian DM de fapt fulfils a contrastive function. All the other values revolve around it and are influenced by it.

4.1.
The core meaning of the Romanian forms is contrastive; non-contrastive usages, with which they sometimes appear, have not been established. Examples (19) and (20) witness contexts where de fapt (to mention only this one) is in fact an elaborative discourse marker. An elaborative discourse marker (Fraser 2009: 296) introduces an utterance that confirms the previous one, plays a key-role in the development of the discourse and sends to an objective reality. All these are features of the Fr. de fait (D'Hondt 2014: 244-246); de fapt belongs to the same semantic field as și ('and', as a DM) and, sometimes, they both appear in texts: de fapt (19) La Ministerul de Interne era de realisat reforma administrativă, și ne vom mândri că am înlăturat pe prefectul numărul doi, care era inutil, căci tot timpul se certa cu celălalt, pentru automobil, pentru onoarea de a figura la cutare ocazie publică. Și de fapt niciodată n-au funcționat doi: ori a funcționat cel care nu era ales, ori a funcționat cel ales. The administrative reform was to be carried out at the Ministry of Interior, and we would take pride in removing the second prefect, who was useless, because he was always arguing with the other one, either for the car or for the honour of being seen at a certain public event. And indeed, they have never worked as a team: either the one who had not been elected fulfilled the tasks, or the elected one. (Iorga,C,16) (20) Pe de altă parte, există nevoia de a se înlătura în favoarea adevăratului muncitor intermediarii arendași, sub forma adeseori primejdioasă și dăunătoare sub care se prezintă azi la noi, și de fapt se manifestă un curent puternic de arendări la obști țărănești.
On the other hand, there is a need to favour the real workers and dismiss the intermediary tenants in the often dangerous and harmful form in which they can be seen in our country today, and indeed there is a strong trend towards peasant common property tenancies. (Brătianu,SC,II,60) Today, de fapt plays a weak role in confirming a prior statement and in preparing a discourse; its occurrence immediately induces a contrastive interpretation, even if there is no contrast in the fragment, as in the following example: According to the newspapers, Austria seems to have a constitution. It seems to have one, we say, because in fact it exists only to be mocked at by a handful of Jews and civil servants who rule this country complex where no one feels satisfied. (Eminescu,P,IX,254) în fapt (23) Apoi, după bunul simț, ne închipuim noi, de aci din țara românească, că au să fie oameni cari să meargă în târg să-și schimbe argintul lor în aur, să plătească un agiu foarte scump poate, numai pentru fericirea ca să meargă apoi la bancă spre al depune acolo? Mare minune! În fapt însă va fi că depunerile la bancă mai le opriți; pentru că în realitate nimeni nu va merge să cumpere aur pentru ca să-l depuie la bancă. Then, as for the common sense, we imagine, from here, from Wallachia, that there will be people who will go to the fair to exchange their silver for gold, to pay a very expensive fare perhaps, only for the short happiness of going to the bank to deposit it there? A true miracle! But, instead, bank deposits will stop; because in reality no one will go buy gold to deposit it in the bank. (Boerescu, DP, II, 1106-1107 de fapt

4.3.
In addition to its basic, contrastive function, de fapt displays in present-day Romanian a reformulation function. This manifests itself in two different ways, related to the intralocutive and the interlocutive dimension of the discourse.
[…] the representatives of the Romanian Police outlined that the origins of this scourge [cases of missing children] lay in one specific problem of the system, actually in two problems of the system, namely extreme poverty and lack of education. (Gheorghe-Dinu Socotar, Cameră, November 24, 2020) In (28) and (29), the speaker actually uses de fapt to correct wordings that he considers inappropriate for the current speech situation; the use of de fapt here implies a perception of the inaccuracies related to the pragmatic aspects of the verbal interaction, such as the observance of the cooperative principle and its maxims or the politeness principle: In (28), Sofia, a TV viewer, talks by phone to Horia Brenciu, the presenter of the morning show 'Neața; after greeting him, she realizes that only "good morning" is not appropriate for that communicative context, because she had been on the show before. On a scale of intimacy with the other person (people who do not know each other → acquaintances → friends → close friends), her initial greeting is appropriate for the first step of the scale, but not for the second, where a greeting should be accompanied by the first name of the interlocutor. In (29), the speaker expresses his desire to meet his interlocutor, but realizes that the modal poate ('maybe') he had just used is too weak, so he restates the proposal, using the sequence trebuie sigur (lit. 'must for sure'), showing a higher interest in the interlocutor and at the same time signalling to the latter that he himself can advance a more precise date for the meeting. In such cases, de fapt seems to fulfil a function similar to elaborative DMs, since it develops and improves an intervention, in terms of pragmatic discourse reality.

De fapt
can actually acquire values that belong to other classes of pragmatic markers, such as topic orientation markers (Fraser 2009: 297) as in (30)

De faptattention marker
At this stage of development, we are interested in all those cases that can no longer be subsumed to the de fapt3 specific structure <S1 -DM + S2>. These new functions, specific to the de fapt4 stage, are gradually reached, and start from the de fapt3 contrastive DM. Consider the fragment: Stop suffocating us with these absurd and totally unethical and illegal questions about the human administration of animal drugs and stop depleting our animal ivermectin stock because of your self-administration! (veterinarian Ovidiu Roșu, on his Facebook page) To give this FB post a proper title, the journalist concisely summarizes the above statement: Ivermectin is an antiparasitic intended strictly for animal use, an utterance that is the representation in the Narrated Discourse 11 (or, more precisely, Narrated Writing) of the above fragment. At the same time, this excerpt shows both the veterinarian's commitment towards the statement "ivermectin is for veterinary purposes only" (which can be explained by introducing this statement though an opinion verb -I consider/believe/am convinced that X), and his opposition to the contrary belief ("ivermectin is also for human use"). The journalist equates the epistemic commitment and the fight against the opposite belief through de fapt. By using this DM he both endorses the opinion of the source and he presents this opinion as the only valid one. The result of the whole process will be as follows: (31) Ivermectina este, de fapt, un antiparazitar destinat strict animalelor Ivermectin is, in fact, an antiparasitic for animal use only (Adevărul, January 26, 2021) Therefore, de fapt belongs to the source (veterinarian Ovidiu Roșu), insofar as it is a Narrated Attitude, i.e. a means of representing a textually manifested state of mind 11 "I will call 'Narrated Discourse' a reference to an enunciative event, without mention of its precise content by way of a completive or an infinitival clause (unlike Indirect Discourse).
[…] As with other S&TP [Speech and Thought Presentation] categories, I will distinguish between Narrated Speech, Narrated Thought, Narrated Attitude and Narrated Writing." (Marnette 2005: 85) through the speaker's firm commitment towards his own statements and by opposition to the contrary opinion. De fapt belongs to the journalist as well, insofar as it is a formula by which the statement and attitude of the source are rhetorically enhanced: de fapt endorses the opinion of the source and thus becomes, from an opinion that the journalist legitimises, the only one that can be maintained.

5.1.
There are two types of de fapt4. The first of these is reached by blurring S1 (the contrasting discourse). This blurring is also a consequence of the fact that the press article where de fapt4 occurs belong to a serial discourse. A serial discourse occurs when there is a sequence of same topic discourses, which may or may not belong to the same author. Specific to serial discourses are, on the one hand, the possibility for each of their elements to refer to any of the other elements, with no specific requirement to specify the serial element that they refer to, and, on the other hand, the fact that any element of the series can reflect and influence the way the reader interprets the facts. The headlines below illustrate a serial discourse: In the press articles headed by the above titles, the reference to a prior discourse is fuzzy, and consequently, the contrastive DM value of de fapt is altered. The process can go one step further, to blurring the source. One common procedure is the weasel word. Weasel words (or anonymous authority) are words and phrases that someone uses to create the impression of a relevant or topical discourse about a certain subject, when, in fact, they make ambiguous and unsupported statements. Weasel words are lexical composites such as: se știe că ('it is known that'), există și opinia că ('there is also the opinion that'), au fost formulate unele critici în legătură cu ('criticism has been made regarding') etc. For example, the title of a press article in Impact is the following: (36) Ce înseamnă, de fapt, când visezi că îți cad dinții What does it actually mean when you dream that your teeth are falling out (Impact, April 10, 2021) De fapt contradicts an opinion not explicitly mentioned within the article and the point of view of the source (also absent) is introduced by anonymous authorities: It is said that if you pay attention to dreams […] they can be used as a kind of portal for your subconscious.
Some say that if you dream that you are losing your teeth, you are suffering from anxiety.
Experts say that dreams in which our teeth fall out are among the most common.
According to studies on dream interpretation, tooth loss is associated with depression.
Under these conditions, the content of the title where de fapt occurs is, in its turn, an anonymous authority. De fapt in the headline makes the reader suppose that (s)he will find within the press article both references to the contrasting previous discourse and to the opinion of the source that the author endorses. These presumptions, triggered by a DM's procedural meaning, are not valid in this context, because de fapt4 is not a contrastive DM. The procedural meaning of de fapt4 differs from de fapt3it functions as an attention marker. In (32)-(36), de fapt displays a rhetorical component in which refutatio vs captatio movements are equally distinguishable. The first suggests that all other information sources hide the truth while only the latter promises to reveal the truth in that article. None of them is actually true: in the first case, refutatio refers to a fuzzy set (the incriminated discourses create a diffuse background), in the second case, the promise is based on a third party's discoursethe sourcewhich is also vague, as the use of anonymous authorities indicates. De fapt4 displays a higher degree of intersubjectivity than de fapt3, because it conveys both the controversy (refutatio, the attitude towards the "other" discourse and its evaluation), and the invitation to read it (captatio, the relationship with the reader).

5.2.
The second use of de fapt4 implies neither a prior discourse, nor an attitude of the source towards it. In these contexts, we deal with de fapt4 as a purely pragmatic, intersubjective and rhetorical marker, displaying exclusively the captatio movement (the function of capturing the reader's attention). This evolution is based on the exploitation of the reading reflexes and of the expectations of de fapt3 standard uses, which refers to a previous S1.
The pragmatic marker that we are talking about probes into the reader's reading habits ("if de fapt occurs, we assume that it refers to a previous element"), but at the same time it deceives these expectations, because the prior discourse that it seems to refer to is actually an empty slot. This use of de fapt4 can be described as follows: 1. its occurrence in the headline creates the expectation of a revelation of the truth (in contrast with another press article, that does not reveal it); but 2. the previous discourse does not really exist; its existence is presumed from the occurrence of de fapt; thus 3. the rhetorical movement of refutatio has no object, is pointless and emphatic, and 4. what remains and matters is only the intersubjective movement of captatio. 5.3. De fapt4 develops into an attention marker, added to the press article title which is the result of a discursive summing up process. Here, de fapt does not contribute to the textual coherence and/or cohesion but it only catches the reader's attention with the promise of an unprecedented revelation. There are at least three distinctions between de fapt3 and de fapt4. We compare the way in which de fapt3 actually embeds S1 and S2 into its definition, <S1 -DM + S2>: 1. the retroactive effect of a DM. De fapt3 introduces S2, but refers to S1this feature was noticed, among others, by Fraser 1999: 942 andRossari 1990: 346, who mentions "the retroactive effect" of DMs. Conversely, de fapt4 refers to a single discourse, S (discourse of the source): (37) Cristina Ciobănașu și Vlad Gherman au spus de ce s-au despărțit, de fapt. Care e motivul real Cristina Ciobănașu and Vlad Gherman revealed why they actually broke up. What is the real reason between this (Fanatik, February 11, 2021) [from the body of the article: "The first day of February arrived with some sad news for the fans of Cristina Ciobănașu and Vlad Gherman. The two actors announced their separation in a six-minute social media video, due to the fact that they evolved differently and no longer shared the same future prospects."] In (37), the headline only refers to the announcement made by the former couple (the title is hence the Narrated Speech of the announcement). Its role of creating a contrast with a previous discourse with a similar topic is null, because there is no such press article. In (37) de fapt does not function as a Narrated Attitude, because it is not significant for the way in which the source -Cristina Ciobănașu and Vlad Ghermanrelates to a previous discourse. This means that de fapt is no longer an amalgam, as in (31), and the journalist is the only one responsible for the occurrence of de fapt in the headline.
The following objections could be made to what has been previously argued: the couple's separation was likely to happen and the reasons behind this were subject to disputes and discussions among their fans; or, this title is likely to contrast with an encyclopaedic content that the journalist implies to be sharing with his reader, content that functions here as a S1. We must keep in mind that these reading presumptions are exactly what the journalist had expected, and their trigger is part of de fapt4 use strategy. This situation is characterized by the fact that a presumption of the pre-existence of a S1 is triggered by de fapt. The title implicitly promises to solve the contrast between an inferred S1 and S2 (opinion of the source), but the press article never fulfils this promise.
2. the time distance between S1 and S2. The formula <S1 -DM + S2> implies a time span between the two segments related by a DM. However, this does not happen in the case of de fapt4. In (38), de fapt4 does not refer to a prior discourse because it cannot do so: the piece of "news" is so recent that there was no time for competing news to come up: (38) Cu ce femeie a petrecut Pepe aseară, de fapt. În ce ipostaze au fost surprinși cei doi Who is the woman Pepe actually spent the evening with. What was the couple spotted doing. (Playtech, February 26, 2021) The refutatio movement, which de fapt enables in (37) and (38), refers to a contentless slot. It is nothing more than an emphatic gesture with no addressee, because it is devoid of object. The two headlines rely exclusively on the effectiveness of the captatio movement.
3. the distinct semantic content of the two segments, S1 and S2, from the DM definition. The difference in meaning between the two statements to which the definition refers is essential, because starting from this point they can be contrastively related. This relationship cannot take place in (39): (39) Ce făceau, de fapt, în Italia doi români care se dădeau drept polițiști sub acoperire What two Romanians who were pretending to be undercover police officers were actually doing in Italy (Digi24, October 25, 2019) The article reveals that the two Romanians mentioned in the title acted deceitfully. Pretending to be a police officer is in itself a deceit. Both the article and its title convey redundant information ('in Italy, two Romanians involved in frauds were actually deceiving people'). De fapt in the title does not have the procedural meaning of a contrastive DM (does not present the news in contrast with a prior discourse).

5.4.
The lack of a previous reference of de fapt4 consequently brings forth word order peculiarities: de fapt4 does not occur in an initial position (position preferred by de fapt3); de fapt4 is embedded either immediately after the predicative verb: "Unde se află, de fapt, Alexandra" ("Where is, in fact, Alexandra"), or it holds the last position in the clause: "Unde se află Alexandra, de fapt" ("Where is Alexandra, in fact"), positions specific for statements with a high degree of subjectivity: "the subjectified elements tend to be positioned peripherally compared to a constituent or a clause" (Traugott 2010: 41).

5.5.
The difference between de fapt3 and de fapt4 is also underlined by the elements that they can be combined with, which more distinctively outline their semantic fields (combinations studied in Fraser 2013 and 2015). Whereas de fapt3 shares the same semantic field with în realitate ('in reality'), de fapt4 belongs to a field that encompasses s-a aflat totul ('everything was revealed') and X rupe tăcerea ('X breaks the silence'). The pragmatic marker de fapt4 and the utterances s-a aflat totul and X rupe tăcerea have identical implicatures: the pieces of information revealed by the press article were secret (or distorted or incomplete) in previous press articles on the same topic:
[in the healthcare sector] total chaos -I had the opportunity to give you an eXAMple. so I went to work one morning↓ and I live in Vitan, in the Potcoavă (horseshoe) neighbourhood and in that green space between the traffic lanes of: # the cars↓ it's a green space↑ and there was an # old man↓ on the ground↓ that morning↓ he fainted and well↓, someone gave him a ride and I told him COME sir (xxx) to the Vitan Clinic and: "WHere do you live, Pops?" (IVLRA: 31) In (42), deci ('so, therefore') is a "narrative cue" (Hoinărescu 2009) that "allows the speaker to indicate a long intervention that follows […] it displays an anaphoric use as a textual deictic and it contributes to the textual coherence and to the integration of the speaker's utterance into a communicational universe, even though not explicitly evoked or described […] the speaker underlines the fact that he took the turn in a conversation and has no intention to pass it on to his interlocutor. The pragmatic inferences that the deci operator triggers, namely the discursive continuity deriving from certain (implicit) premises, increase its perlocutionary efficiency and grant the features of a metadiscursive signal such as 'Attention!', 'I'm not done!', 'I don't want to be interrupted!'" (Hoinărescu 2009: 51) De fapt4 displays similar characteristics as deci. It indicates to the reader the details to be presented in the press article, it acts like a textual deictic (without being a real one) and it requires the reader's attention by an implicit promise that something worthy of attention follows. Usages as (42) obviously differ from (43), where deci is an inferential DM (Fraser 2009: 296- Therefore, it is no by chance that de fapt and deci developed into attention markers in Romanian. With this function, they achieve their goal of drawing attention to the importance of the following utterances by making the best use of the traits that they share with the corresponding DM. Examples (37)-(41) are not isolated. Similar headlines are released daily, and de fapt has become a reliable means of catching the reader's attention and of encouraging the latter to continue reading the press article. De fapt4 titles are frequent today in publications such as Playtech, Impact, Capital, Voxbiz or Adevărul. A Google search of de fapt in the digital archives of Playtech magazine, for example, returns 23 pages, with 330 results in recent years. The first title in Playtech in which de fapt appears as an anonymous authority can be dated back to June 13, 2011: "Apple a creat de fapt Androidul" ("Apple actually created Android"; the first line of the article contains a weasel word: "This is not a personal opinion, but I must say that I find it extremely interesting"). A similar search, performed in the Impact archives, returns 60 de fapt titles for articles published between March 27 and April 16, 2021, i.e. an average of 3 per day (for this publication only). Even if de fapt does not display the features outlined above everywhere in these titles, its frequent use ultimately determines its migration to the stage that we have discussed in this study.
The current uses of de fapt (rhetorically promoted in press headlines) and deci (frequently used in common Romanian) clearly indicate a pragmatic shift to the category of attention markers.

Conclusions
The evolution of the Romanian form de fapt, which follows exactly the three stages Elizabeth Traugott hypothesised, can provide a glimpse of what happens when the adverbs that are part of this pattern accomplish the transition from stage 3 to stage 4.
The first relevant thing is that for the description of de fapt4 the DM general scheme <S1 -DM + S2> can no longer be used. Thus, the pragmatic mark de fapt4 loses a common feature of the previous stages, namely the semantic connection with a prior discourse. Consequently, the only appropriate description is S + DSM.
Through this discursive behaviour, de fapt moves onto a new category of pragmatic markers, that of discourse structure markers, a category that also includes iată ('look'), uite ce voiam să spun ('here's what I wanted to say'), să vezi ce s-a întâmplat ('you won't believe what has happened'), hai să-ți spun ceva ('let me tell you something') or deci ('so, therefore') in (42). The key milestones in the evolution of de fapt towards this stage are: 1. the loss of the connection with the previous discourse and its use as an anonymous authority; 2. the use of its polemic potential, inherited from the corresponding DM, which, in the case of de fapt4, becomes a void rhetorical gesture (it refers to an empty slot, with a missing previous discourse); 3. the use of a reading reflex, based on the sine qua non condition of DMs: if de fapt3 occurs, it must refer to a previous discourse. The use of de fapt4 in a title is a promise of revealing a hidden truth, provided that the reader goes through the content of the article. The stake of this context is the invitation to read the content and not to emphasize a contrast with a previously formulated point of view, which, in the case of de fapt4, is just a pretext. Thus, de fapt becomes the easiest way to make a story and catch people's attention.
Therefore, at this stage, the journalist-reader relationship plays a key role. In other words, de fapt4 acquires, for the first time ever, an intersubjective nature.
The frequency with which de fapt is used as an attention marker in press articles is likely to indicate either a successful recipe or its ending productivity. The answer can only be given following a more thorough research, in which we should compare, among other things, the number of clicks a piece of news with a de fapt headline receives as opposed to a headline where de fapt is absent. Until then, we can only restate Peter Blumenthal's comment: "the phrase en fait has become a cliché for so many speakers, being used every two sentences." (Blumenthal 1996: 266).