1 Fieri in Daco-Romance as a case-study in “standard language bias”

It is generally held that Latin fieri ‘be made, become, happen’ is today extinct in all branches of the Romance languages except for Daco-Romance. In the latter, reflexes of fieri provide some parts of the inflexional paradigm of the verb ‘to be’, particularly the subjunctive, the infinitive, and the gerund, while forms historically derived from Latin esse ‘to be’ occupy the rest of the paradigm. The morphological history of fieri within the Romanian verb ‘to be’, and particularly in the subjunctive, has been surprisingly neglected in histories of the Romanian language (cf. the paucity of discussion of the topic in, for example, Densusianu 1938; ALR volume VII; Coteanu 1969; Zamfir 2005; Sala/Ionescu Ruxăndoiu 2018). In what follows I shall examine that history more closely and offer comparative-historical evidence that, in the earliest forms of Daco-Romance, fieri had a different paradigmatic distribution from that of modern Romanian. I will argue that these findings about the history of this phenomenon should stand as an example of how our understanding of the mechanisms of language change is liable to be distorted by what I call “standard language bias”, the tacit assumption in reconstruction according to which structures familiar to us from the standard languages which we tend to know best are somehow assumed to be “basic” and “normal”. I will also sketch some other examples from Romance languages of the undesirable effects of such “standard language bias”, showing how it can lead to misunderstandings of the history of the standard languages themselves, as well as of the history of their cognate varieties.

To appreciate the diachronic development discussed in my principal case-study, it is first necessary to review the place of forms derived from fieri^[1] in the verb ‘to be’ in each of the four branches of Daco-Romance, namely Daco-Romanian (the group of dialects which includes standard Romanian), Istro-Romanian, Megleno-Romanian, and Aromanian. It will become apparent that these varieties display a high degree of similarity in the paradigmatic distribution of fieri, but also that there are very significant differences.

2 Reflexes of fieri and their morphological distribution in the verb ‘to be’. A cross-dialectal synopsis

Table 1 shows the distribution of fieri (marked in bold) in the modern standard Romanian verb ‘to be’.

All Daco-Romanian dialects share at least this basic paradigmatic distribution of fieri although, as we shall see, in some varieties fieri also occurs in other parts of the paradigm.^[2] The position in Istro-Romanian, Megleno-Romanian, and Aromanian is summarized in Tables 2, 3, 4. Each of these branches of Daco-Romance shows extensive variation among subvarieties for the verb ‘to be’, so that in certain respects the material given in the tables overlooks more complex linguistic realities. The sign “?” beside a form indicates that either its etymology or its function as a present indicative is debatable; these cases will be discussed later. Examples for Istro-Romanian^[3] (Table 2) are mainly based on data gathered by Pușcariu (1926, 196):

The first and second person plural forms fim, fiʦ (cf. Romanian fim, fiți) are heard in the dialect of Žejane (Kovačec 1998, 291), and are clearly historically older forms. For Megleno-Romanian (Table 3), we may take the data provided by Capidan (1925, 172–74). In this variety *[fi] > [i]:

Capidan specifies that the first person singular subjunctive form ([iu̯]) is mainly used in the dialect of Huma. Atanasov (2002, 250) shows, for the first and second persons singular of the subjunctive, only forms that are also those of the present indicative. Capidan (1932) also provides an authoritative overview for Aromanian (cf. also Weigand 1888, 99–100.; Boiagi 1915; Papahagi 1932, 78–79; Saramandu 2007, 444; Nevaci 2011, 160). Here, too, there is very considerable dialectal variation, only part of which I can render below (Table 4); my focus is particularly on those parts of the paradigm in which fieri persists. Note that here *[fi] > [hʲi]):

The picture for Daco-Romance as a whole can be summarized as follows. Continuants of fieri are found in all varieties in the subjunctive, imperative, infinitive, and gerund.^[6] In Megleno-Romanian and Aromanian, the first and second person plural present indicative are exclusively continuants of fieri and, in Aromanian, forms of fieri may also appear, at least in some dialects,^[7] in the first and second person singular of the present indicative. Istro-Romanian intriguingly presents some evidence, albeit fragmentary, for fieri in the present as well as the subjunctive. Neiescu (2015, s.v. fi) records examples such as Țela istu cårlefii̯eîn voz hitɛ firu ‘That very person who is in the cart is throwing the hay’, or Fii̯e råče ‘It is cold’.^[8]Cantemir (1959, 166) gives Și fiiu hurușvele dosti bure in fact translated as ‘Și perele se fac foarte bune’ = ‘And the pears are becoming very good’. Popovici (1909, 14), in a text from Šušnjevica, gives I̯o voi̯ ie zis-au̯ se reș samo ști i̯uve vo̯ fiieți ‘I will, he said, if only I knew where you are’. Weigand (1892, 246) also presents a possible example.^[9] Istro-Romanian is the only variety in which fieri also appears in the imperfect, and as a variant form of the past participle. Istro-Romanian marks aspect even in the past participle, and the alternative past participle form [fiˈjei̯t] (Weigand 1892, 248; Streller 1904, 31; Popovici 1914, 77) shows creation of an imperfective form of the past participle derived from the infinitive and combined with the imperfect indicative marker [ei̯]. The type [fiˈjejam] may be a purely local creation of a past imperfective form of the verb ‘to be’, combining the fi- root with the imperfect tense endings found in other verbs.

3 What the comparative Daco-Romance data really tell us

These data suggest that in the proto-language a form having root [fi]- (i.e., the derivatives of fieri) was found in the infinitive, the gerund, and the subjunctive. But the data from Aromanian and Megleno-Romanian, and possibly also those from Istro-Romanian, significantly further suggest that the [fi]- root might also originally have been found in the present indicative, particularly in the first and second person plural forms thereof. I believe this suggestion to be not only plausible but most probably correct. However it is not adopted by historians of Daco-Romance: the facts tend to be viewed, rather, through what might be seen as the distorting lens of modern Romanian, where, among finite forms of the verb, fieri is of course restricted to the present subjunctive (with the imperative).^[10] Thus Streller (1904, 5; 11) says that Aromanian 1sg.prs.ind [hʲiu], 1pl.prs.ind [hʲim], 2pl.prs.ind [hʲiʦʲ] are originally subjunctive forms, which have been analogically extended from the subjunctive into the present indicative. He takes Romanian to represent the historically underlying situation, while “[i]m Aromunischen haben wir in der 1. plur. nur hʲim[u aus lat. fimus, welches im Rumän. eigentlich die Form des Konjunktivs ist und ursprünglich wohl auf diesen beschränkt war” (Streller 1904, 11). Streller makes the same analysis for Megleno-Romanian. Similarly, Capidan (1925, 173) asserts that Megleno-Romanian 1/2pl.prs [im]/[its] are “împrumutate de la conjunctiv, ca în dialectul aromânesc”, while Capidan (1932, 488) also says of the Aromanian1sg.prs variants [hʲiu̯], [hʲim] that “[o]riginea lor este clară dela conjunctiv”. Rosetti (1986, 148) reiterates Capidan’s interpretation.

The idea that these present indicative forms of fieri come from the subjunctive is actually back-to-front. Rather, the modern subjunctive forms must come from the present indicative of fieri, at least in the first and second persons. Most scholars, clearly influenced by modern Romanian, seem to assume that because the fieri forms are today characteristic of the subjunctive in the standard language, they are inherently, and have always been, subjunctive forms – and consequently that they come directly from the Latin present subjunctive series fiam fias fiat fiamus fiatis fiant: for this assumption cf., for example, Philippide^[11] (2011, 479); Bourciez (1923, 541); Capidan (1925, 174); Rothe (1957, 114); Rosetti (1986, 148). It is insufficiently appreciated that Romanian first and second person fiu fii fim fiți are simply not phonologically possible reflexes of fiam fias fiamus fiatis,^[12] for from these we should have expected something like **fie **fie **fiem **fiați. The series fiu fii fim fiți is, by contrast, impeccably derivable from Latin present indicative fio fis fimus fitis. In short, the first and second person forms, singular and plural, of the Romanian present subjunctive must be originally present indicative forms. Where the same forms also appear in the present indicative, as they consistently do in Megleno-Romanian and Aromanian, they are no more or less than what they appear to be: not imports from the subjunctive but the regular continuants, in situ, of the Latin present indicative of fieri. As for the third person subjunctive form fie (which is both singular and plural), this is a phonetically regular reflex both of Latin 3sg.prs.sbjv fiat and of 3pl.prs.sbjv fiant. While I think it most likely that fie, both singular and plural, does derive from the Latin subjunctive forms, even this is not entirely certain: the third person singular could also be an original indicative. Latin indicative fit would be expected to yield **fi in Romanian, not fie. However, if we compare this with the fate of Latin scit ‘s/he knows’, which in modern Romanian has yielded not expected **ști^[13] but știe – showing analogical introduction of the inflexional -e characteristic of other non-first conjugation third person singular present verbs – we see that fie is also a possible original present indicative form. In this case, there would also be a plausible explanation of plural fie as an originally indicative form, simply because there is always total syncretism in Romanian between the third person singular and third person plural subjunctive. Plural fie could, conceivably, constitute a simple replication of that pattern of syncretism, merely “copying” the corresponding singular fie.^[14]

In general, the difficulties of deriving the full set of Romanian present subjunctive forms from Latin subjunctives are not properly addressed in the literature. When they are considered, this is done in a curiously ad hoc manner, as if ‘to be’ were somehow an entirely special case,^[15] isolated from the rest of Romanian verb morphology. Thus Rothe (1957, 114) seeks to account for the phonological discrepancy with the supposed underlying Latin forms by invoking a special analogical introduction of -u into the first person singular to replace expected -e (cf. also Capidan 1925, 174 for Megleno-Romanian). To explain the forms of the first and second persons plural, however, he invokes the analogy of the present indicative forms of fourth conjugation verbs, without offering any plausible account of why the analogy of the fourth conjugation would operate just in these cells.

Only one person,^[16] namely Friedrich Streller, has ever spelled out a truth which I believe to be self-evident, namely that the first and second person forms (at least) of the Daco-Romance subjunctive of the verb ‘be’ originated as present indicative forms and were subsequently introduced from the present indicative into the subjunctive. Streller (1904, 17) writes: “Von diesen Formen [scil. those of the subjunctive] geht nur die 3. sing. und die 3. pl. lautlich auf die lateinische Konjunktivform zurück [...]. In den übrigen Personen sind wie bei allen anderen Zeitwörtern die Formen des Indikativs eingedrungen”. Streller’s interpretation seems to have been ignored by historians of Romanian morphology, but it has the great virtue of situating ‘to be’ in its wider context, namely as being subject to the generalization that all Daco-Romance verbs show historical replacement of Latin present subjunctive forms by continuants of the Latin present indicative in the first and second persons singular and plural.^[17] This situation is exemplified in Table 5.

Romanian subjunctives are indeed all systematically distinguished from the present indicative, but this is done by means of the subjunctive-marker să which normally precedes the verb (e.g., 1sg.prs.indcânt, vând, dorm vs 1sg.sbjvsă cânt, să vând, să dorm; 3sg.prs.indcântă, vinde, doarme vs 3sbjvsă cânte, să vândă, să doarmă).

The Romanian third person subjunctive forms certainly continue their Latin third person singular and plural present subjunctive ancestors (e.g., cânte < cantet/cantent, vândă < uendat/uendant, doarmă < dormiat/dormiant),^[18] but the other Romanian subjunctive forms cannot directly continue Latin present subjunctives. For example, Latin first person singular present subjunctive cantem, uendam, dormiam would regularly have yielded Romanian **cânte, **vândă, and probably **doarmă, while Latin second person plural present subjunctive cantetis, uendatis, dormiatis would have yielded **cânteți, **vândați, and probably **dormați. Rather, the first and second person singular and plural forms of the Romanian subjunctive clearly continue Latin present indicative forms: e.g., cânt < canto, vând < uendo, dorm < dormio;cântăm < cantamus, vindem < uendimus, dormim < dormimus. Indeed, for all Romanian verbs other than ‘to be’, it is generally accepted, even by those who would treat the subjunctive morphology of ‘to be’ as a special case, that there has simply been wholesale replacement of the subjunctive by the present indicative: thus, for example, Philippide (2011, 499); Rothe (1957, 90); Rosetti (1986, 144). In fact, in trans-Danubian Daco-Romance the tendency is for the subjunctive forms to be wholly replaced by present indicative forms, especially in the first conjugation, whilst in Istro-Romanian this is true for all verbs except ‘be’ (Capidan 1925, 161; 1932, 445; Pușcariu 1926, 196; Kovačec 1971, 150). Hurren (1999, 104) detects signs that in Istro-Romanian the distinctive subjunctive form is being ousted by the indicative even in the verb ‘to be’. In northern Aromanian varieties the first person singular form is generally of the type 1sg.sbjv [ˈesku] < 1sg.prs [ˈesku], 2sg.prs [ˈeʃti] while in Megleno-Romanian we have alternative 1sg.sbjv [sɔm], 2sg.sbjv [jeʃ], which correspond to the present indicative forms rather than to fieri.^[19]

In sum, it is counterintuitive, and contrary to other observed developments, that a subjunctive form would oust an indicative, in the verb ‘to be’ or in any other.^[20] Overall, the analogical replacement of maximally “unmarked” present indicative forms by the more “marked” subjunctive seems unlikely, and there are, to my knowledge, no examples of this happening systematically in other Romance languages. All the evidence shows that the present indicative of fieri survived in early Daco-Romance, at least in the first and second persons. The relevant present indicative forms continue to this day in Aromanian and Megleno-Romanian, but they have been lost in Daco-Romanian varieties.^[21] Their original existence is further presupposed in all Daco-Romance varieties by the fact that the first and second person forms of the subjunctive, of all verbs, can be shown to have originated as present indicatives.

This leaves various unanswered questions, only some of which I am at present able to resolve. One is whether early Daco-Romance also had continuants of fieri in the third person present indicative. Another is how fieri was originally related to the continuants of esse, and what the distinction between the two verbs, if any, was. A related question is why fieri replaced esse in specific parts of the paradigm, and vice versa. For some answers to these questions it will first be helpful to review some comparative evidence from other Romance languages.

4 Fieri in the wider Romance perspective

The notion that early Daco-Romance once had reflexes of fieri outside the present subjunctive and the infinitive and the gerund is well supported by comparative evidence from other Romance languages,^[22] and particularly from medieval northern Italo-Romance and Tuscan, where the reflex of fieri was widely used in the present, imperfect, and future, and in both the indicative and the subjunctive. It is very well known (cf., e.g., Rohlfs 1968, 272; 1969, 129–130; Michaelis 1998; Cennamo 2003), that in these Italo-Romance varieties the reflex of fieri served as the auxiliary verb in passive constructions comprising ‘be’ + past participle. What is scarcely ever observed (Michaelis 1998, 81 is an exception) is that fieri also survived extensively outside passive constructions. In the north, in addition to its function as a passive auxiliary, it preserved its original meaning of ‘become, come about, come to be, happen’. Some examples of present indicative forms from the early fourteenth century Milanese Volgarizzamento in antico milanese dell’Elucidario (data from OVI database) are:

1. Doncha nuyfimocrucificadi con Criste al mondo [...]; e in lo batessimofimosepelidi con Criste e imperzò nuyfimosomerzuy trea fiada soto l'aqua per significare li tri dì de la morte de Criste; per la vivanda del So corpo nuyfimouna medexima cossa con Criste

‘So we are crucified with Christ in the world [...]; and in baptism we are buried with Christ and therefore we are immersed three times under water to signify the three days of the death of Christ; by consuming His body we are/become one with Christ.’

2. Sì como la testa è zonta al corpo e per la testa se reze lo corpo, in-cossì la Giexa è conzonta a Criste per lo sacramento de Criste, zoè per lo corpo So, efinuna cossa con Luy, da lo qua tugi li iusti in so grado e in so ordenefinrecevudi e governadi sì como membri da lo cò.

‘Just as the head is joined to the body and the body is governed through the head, so the Church is joined to Christ through the sacrament of Christ, that is through His body, and they are/become one thing with Him, by whom all the just according to their rank and order are received and governed like limbs by the head.’

3. E di' savere che, poy ke l'omo è batezado, elfitemplo de Deo, sì com dixe santo Polo.

‘And you must know that, after a man is baptized, he is/becomes a temple of God, as Saint Paul said.’

4. Quilli ch'in [...]inferno san le cosse chefinin questo mondo in quella guixa che nuy vivi savemo chefiin inferno

‘Those who in hell know the things that are/happen in this world in the way that we the living know what is/happens in hell’

In medieval Tuscan, the reflexes of the Latin present indicative of fieri are also well represented. They do not seem to preserve the meaning ‘become’, as in the north, but they tend to serve as future-tense forms of the verb ‘be’ (Rohlfs 1968, 336). Again, these forms are not restricted to passive constructions. An example from the fourteenth century Breve dell’Arte de’ calzolai di Pisa (data from OVI database) is:

5. e catuno di loro, per buona fede, sensa fraude, allo meglio ch'elli sapràve, di tutte quelle cose unde li domandràe consillio per lo fatto dell' arte; et di venire a consillio quando richiestofie^[23]

‘and each of them, in good faith, without fraud, as best as he would know, of all those things about which it will ask him for advice about the business of the guild; and to come to counsel when he is/will be asked’

6. et ad honore di messer la Podestà del Comune di Pisa, et di messer lo Capitano et de li signori Ansiani del populo di Pisa, che ora sono etfinnoper li tempi

‘and in honour of messer the Podestà of the Commune of Pisa, and of messer the Captain and of messers the Elders of the people of Pisa, who are now or will be at the time’

While medieval Italo-Romance preserves the best-known examples of the survival of present tense forms of fieri, it is not generally recognized that at one time Sardinian also had such forms.^[24] Here, what were originally present indicatives of this verb have, rather curiously, ended up not as present indicatives at all but as preterites – or sometimes even as imperfects – of the verb ‘to be’. Wolf (2014, 29) is apparently the only person to have realized that “[d]ans une grande partie du nord de la Sardaigne,^[25] le présent de fieri s’est [...] substitué au prétérit de esse”. However, Wolf argues persuasively that preterite forms such as [ˈfippo] [ˈfizi] [ˈfidi] [ˈfimus]/[ˈfimis] [fidzis] [ˈfini] are explicable as being original present indicatives continuing fio fis fit, etc., and that they are not (as suggested for example by Wagner 1939, 18) reflexes of Latin perfective fui fuisti fuit, etc. Wolf (2014, 26–27) does not have a particularly satisfactory account of how those original present forms ended up as preterites, but it seems to me that a combination of semantic similarity and strong segmental and prosodic resemblance with the inherited preterite of the verb ‘be’ in fu- (compare preterite fit/fidi < prs.ind fit with preterite fut/fudi < pst.prffuit elsewhere in Sardinian) must have prompted reanalysis of what was originally a present indicative of fieri as a preterite of the verb ‘to be’.^[26] The failure to see that the relevant preterite or imperfect forms are actually old present indicatives in Sardinian rather echoes linguists’ reaction to the paradigmatic distribution of fieri in Romanian: because the original present indicative forms appear in a part of the inflexional paradigm other than the present indicative, there has been reluctance to recognize that they are, historically, present indicatives.

5 Fieri in early Daco-Romance

It is notoriously difficult to reconstruct the grammatical system of Daco-Romance in the first millennium, but the wider Romance context briefly described above licenses educated guesses. Certainly, the notion that early Daco-Romance had a set of present indicative (as well as of subjunctive) forms derived from fieri, as suggested by the “internal” comparative evidence of the trans-Danubian dialects, is “externally” supported by northern and central Italo-Romance, where such forms survive in their original function, and by Sardinian where there is evidence that they once existed. But what was the meaning of the continuants of fieri in early Daco-Romance? Were they already synonymous with the continuants of esse? Here both the external and the internal comparative evidence points to the survival of what is sometimes called “fientive” (or “ingressive”, “transformative”) meaning. We have seen that this meaning persisted into old northern Italo-Romance, and such a meaning provides clear motivation for the specialization there of the verb as a passive auxiliary (cf. the use of German werden ‘become’, or English get, as passive auxiliaries), probably originating in resultative expressions. In both Tuscan and Vegliote this verb shows a particular association with the expression of future meaning. In Tuscan, it is the continuants of the Latin present indicative (and subjunctive) that acquire the meaning ‘will be’. Moreover in Vegliote this verb (Tekavčić 1977, 71–77) survives solely in a future tense form ([ˈfero]), and constitutes the sole, and suppletive, future form of the verb ‘to be’.^[27] I suggest that the particular link between surviving forms of fieri and future meaning in both Tuscan and Vegliote is not accidental, but consistent with the persistence of the meaning ‘become’ in early Romance. Consider the following propositions:

1. The sea is becoming warm.

2. The sea will be warm.

3. The sea will become warm in the summer.

4. The sea will be warm in the summer.

Proposition (A) strongly implies (A´), and proposition (B) inescapably entails (B´). In fact, there are many circumstances in which (B) and (B´) could be effectively synonymous. Given this, it is not surprising that the present tense of ‘become’ should have been susceptible to reanalysis as having inherently future meaning, as in Tuscan, or that the future of ‘become’ could be analysed as a synonym of the future of ‘to be’, thereby potentially replacing the latter. For more detailed discussion of the role of synonymy or near-synonymy in generating suppletion, cf. Maiden (2018, 206–300). The crucial point is that the Tuscan and Vegliote facts make good sense just if we assume that fieri still preserved in early Romance the meaning of ‘become’. This conclusion is at least indirect evidence that the meaning ‘become’ probably endured in old Daco-Romance and it is a conclusion supported by the internal evidence of Daco-Romance. There have been few attempts to explain the intraparadigmatic distribution of fieri in Daco-Romance, but Lombard (1955, 727) appeals to what he calls a “nuance futurale” in the infinitive and subjunctive which might have favoured their replacement precisely by a verb originally meaning ‘become’.^[28] I think Lombard’s basic insight is correct, although I would frame it slightly differently. Consider the following Romanian examples:

C. Vreau ca marea să fie caldă.

I.want comp sea.the comp be_sbjv warm

‘I want the sea to be warm.’

C´. Vreau ca marea să devină caldă.

I.want comp sea.the comp become_sbjv warm

‘I want the sea to become warm.’

By their very nature, verbs of “becoming” express the realization of something not yet realized (a fact which also favours the reanalysis of such verbs as future auxiliaries): only something which is not yet warm can become warm. Romanian subjunctives are irrealis forms, par excellence, so that the subjunctive of a verb meaning ‘become’ and the subjunctive of a verb meaning ‘be’ will, in many contexts, be practically or wholly synonymous, as examples (C) and (C´) illustrate.^[29] In both cases it is implicit that the sea is not now warm, but that the subject wants it to be(come) so. The frequent overlap between ‘being’ and ‘becoming’ in irrealis contexts perhaps facilitated the suppletive replacement of the former by the latter. The same kind of reasoning may be applied to the infinitive which, as Lombard points out, is frequently employed in irrealis contexts (e.g., pentru a fi ‘in order to be’, fără a fi ‘without being’, poate fi ‘it can be’, or indeed in future tense constructions which originally comprised the verb ‘want’ + infinitive, as in va fi ‘it will be’). The crucial point, again, is that the modern paradigmatic distribution of fieri in Romanian makes good sense historically particularly if we recognize that, in early Daco-Romance, the sense of ‘become’ still persisted. Finally, Cantemir (1959, 165–166) says that Istro-Romanian fi can mean ‘happen’, giving the example Se nu re fi ceva slabo ‘Let nothing bad happen’, while he translates the sentence Și fiiu hurușvele dosti bure as ‘Și perele se fac foarte bune’ = ‘And the pears are becoming very good’, which indeed seems to be the most plausible rendering in the context.

The foregoing discussion actually renders the situation found in Aromanian and Megleno-Romanian (and perhaps in Istro-Romanian) rather puzzling. If the reflexes of fieri originally retained the meaning ‘become’, we may understand why they have emerged as subjunctives and infinitives of the verb ‘be’, as just explained, but why would they have replaced the reflexes of Latin esse in the present indicative, and particularly the first and second person forms? After all, ‘we are’ and ‘we become’ are not obviously synonymous. The problem may need to be approached from a different angle. In Aromanian and Megleno-Romanian fieri is the only form found for ‘we are’ and ‘you (plural) are’. Now in Daco-Romance generally the first and second person plurals of the present of the verb ‘be’ are an area of complete discontinuity with the corresponding forms of Latin esse: nowhere does any reflex of Latin sumus and estis survive. In old Romanian the relevant cells of the paradigm were sometimes filled by reflexes of late Latin present subjunctive sĭmus sĭtis (cf. note 20 for discussion of these forms and their bearing on the present study). In most modern Daco-Romanian dialects (and in the Aromanian dialect of Siracu – Capidan 1932, 484) we find the kind of outcome represented by standard Romanian suntem sunteți, which shows analogical generalization of third person plural sunt. In Istro-Romanian, these cells are filled by forms borrowed from Croatian (smo, ste). So while we cannot be sure why Aromanian and Megleno-Romanian deployed reflexes of fieri, such behaviour is not surprising in the wider context of a tendency to fill these two paradigm cells by importing extraneous material, combined with the fact that in the subjunctive and the infinitive fieri had suppletively replaced esse and that the semantic distinction was therefore lost in those parts of the paradigm. A similar explanation may be available for the first and second person singular present reflexes of fieri in some Aromanian varieties. In Romanian and many Daco-Romanian dialects, the predicted reflexes of Latim sum, namely [su], [ɨs], or [s], have been replaced by the originally third person plural form sunt. In Istro-Romanian and Megleno-Romanian the forms [sǝm], [sam] are borrowed from Croatian and Macedonian respectively. In Aromanian, the expected continuant of sum appears only to survive in Fărșerot dialects ([ɨs], -[s]); elsewhere we generally have the type [esku], analogically modelled on second person singular [eʃtʲ]. In short, and perhaps surprisingly, the first person singular present indicative of the verb ‘to be’ in Daco-Romance turns out to be a frequent “weak point” in the paradigm of the verb, and one liable to be filled, again, by extraneous morphological material. I suggest that it is in this context that the appearance of the form [hiu] ‘I am’ in some Aromanian dialects should be viewed.^[30]

6 The perils of “standard language bias”: some more examples

The evidence that we have seen suggests that in early Daco-Romance the reflexes of fieri survived not only in the subjunctive (and in the infinitive and gerund), as in modern Romanian, but also in at least the present indicative. The modern distribution of continuants of fieri within the paradigm of the Daco-Romance verb ‘to be’ is also consistent, for the reasons just explained, with the notion that in early Daco-Romance the reflexes of esse and those of fieri were not yet perfectly synonymous, the latter retaining connotations of ‘becoming’. These conclusions seem to me little more than a statement of something which should always have been obvious. They are remarkable only in that, to the best of my knowledge, nobody has ever said them before. The reasons for this omission carry an important methodological message: they reveal the potentially distorting effects of allowing familiarity with a well-known, standardized, linguistic variety to condition one’s comparative and historical analysis. To properly understand the history of the verb ‘to be’ in standard Romanian one needs, then, to regard Romanian as just one piece in the comparative jigsaw puzzle, not the whole picture. I further suggest that this perspective should be extended to the historical study of languages more generally. We should keep a critical eye open for the possibility that our analyses are unduly conditioned but our familiarity with standard languages.

As illustration that such “standard language bias” is by no means limited simply to the Romanian verb ‘be’, I offer three brief sketches from the history of Italo-Romance (one of which also involves Romanian). I stress that my intention here is not to regurgitate the details of arguments which I or others have made elsewhere, and which readers can evaluate for themselves from the references, but simply to indicate how linguists may be influenced by “standard language bias”. The first example involves approaches to rafforzamento (or raddoppiamento) fonosintattico in Italian, that phenomenon whereby under certain circumstances word-initial consonants are lengthened: e.g., io [p]arlo ‘I speak’ vs tu [pː]arli ‘you speak’; canto [b]ene ‘I sing well’ vs cantò [bː]ene ‘she sang well’; ogni [ʤ]orno ‘every day’ vs qualche [ʤː]orno ‘some days’. Loporcaro (1997) argues, in effect, that approaches to the history of the phenomenon have suffered from a perspective which projects into the past the position obtaining in modern standard Italian, such that rafforzamento appears to be closely (but not entirely) correlated with the phonological environment of a preceding stressed vowel, inviting prosodically-based explanations which appeal to the interaction of stress with syllable-structure constraints. Loporcaro takes a historical-comparative perspective which considers in detail the totality of Italo-Romance (and particularly southern Italo-Romance) dialects, from which the conclusion persuasively emerges that rafforzamento fonosintattico in standard Italian, as in the dialects, is in origin a matter of consonantal assimilation across word boundaries. The environment for this assimilation was later opacified, favouring analogical spread of the phenomenon to new environments in which the original assimilation did not originally operate. Loporcaro characterizes many then current approaches to the history of rafforzamento fonosintattico as follows (Loporcaro 1997, 39):

Le recenti proposte [...] di spiegazione diacronica del RF sono basate sul procedimento della ricostruzione interna: costatando condizioni oggi valide per l’italiano [...] se ne opera una proiezione su di una fase anteriore dello sviluppo della stessa varietà per ricavarne un’ipotesi circa l’evoluzione diacronica dei fenomeni in questione.

Loporcaro further argues (1997, 40) that

ogni ipotesi sull’origine del RF in italiano, formulata in base a dati interni, dev’essere passata da un lato al vaglio della documentazione diretta disponibile seguendo prospetticamente nelle sue attestazioni scritte il percorso diacronico che conduce dal latino all’italiano standard a base toscana. E, d’altro canto, dovrà essere confermata dalle argomentazioni ricostruttive che è possibile sviluppare in forza degli elementi di prova offerti dalle altre varietà romanze.

Loporcaro’s monograph is, in fact, an excellent example of the superior power of such an approach.

Another example concerns the history of diphthongization in standard Italian (and in Tuscan dialects). The diphthongs ie [jɛ] and uo [wɔ] originate from the historically underlying vowels [ɛ] and [ɔ] in stressed open syllables.^[31] They remain restricted to stressed open syllables (leaving aside some cases of analogical extension): e.g., [ˈmwɔ.re] muore ‘(s)he dies’ but [ˈmɔr.to] morto ‘dead’, [ˈtjɛ.ne] tiene ‘(s)he holds’ but [ˈtɛŋ.go] tengo ‘I hold’. In many Italian dialects outside Tuscany, in contrast, we find similar diphthongs, also in stressed syllables, but with an otherwise different distribution, since they occur equally in open and in closed syllables, and are limited to a particular historical phonological environment. They are the product of metaphony, an assimilatory process triggered by original following unstressed [i] or [u]: e.g., central Italo-Romance [ˈmɔ.re] ‘s/he dies’, [ˈmwo.ri] ‘you die’, [ˈmwor.tu] ‘dead_msg’, [ˈmɔr.ta] ‘dead_fsg’; [ˈtɛ.ne] ‘s/he holds’, [ˈtje.ni] ‘you hold’, [ˈtjeŋ.gu] ‘they hold’, [ˈtɛŋ.go] ‘I hold’. The prevailing tendency (for example, Castellani 1980) is to assume that there is no connexion between the Italian diphthongization, on the one hand, and metaphonic diphthongization, on the other. Indeed, metaphony is conventionally assumed not to have operated in the history of Tuscan. The notion that the Italian diphthongs and the dialectal metaphonic diphthongs could be historically connected has even been dismissed as a “myth”.^[32] Italian, together with the Tuscan dialects from which it derives, tends to be treated as a ‘special case’, and divorced from the metaphonic diphthongization found in other dialects. In contrast, some scholars (cf. Maiden 2016 for an account of the history of the question) have argued, on the basis of detailed comparative analysis of the two kinds of diphthongization, that there is in fact an intimate historical connexion between them. Maiden, for example, points to evidence that, (a), metaphonic diphthongization, like modern Italian diphthongization, may originally have been restricted to closed syllables and, (b), that exceptions to expected diphthongization in modern Italian and in Tuscan dialects correlate significantly with the absence of a historical metaphonizing environment ‒ a fact that suggests that where the diphthongs do occur they were originally associated with metaphony. If Maiden and others are correct, then Tuscan and Italian turn out to be a lot less ‘special’, when compared with other Italo-Romance dialects, than is often assumed. The important point here is that proper understanding of the vexed question of the origins of Tuscan diphthongization can only be attained by considering Tuscan (and Italian) as just one Italo-Romance variety among many.

My final example of the perils of “standard language bias” brings us back to Romanian, but crucially also involves questionable assumptions about standard Italian. Elson (2017) argues that a pattern of allomorphy in the present tense of Romanian verbs is due to contact with “middle Bulgarian”. For the details and their theoretical ramifications see both Elson (2017) and Maiden (2021). Simplifying somewhat, Romanian fails to show an etymologically expected, phonologically regular, form of allomorphy in the third person plural of verbs. While both Romanian and Italian show the effects of the relevant sound change in the first person singular and in the present subjunctive (e.g., uenio ‘I come’ > *[ˈvɛnjo] > *[ˈvɛ(ɲ)ɲo] > OIt. vegno, ORo. viu; ueniat ‘(s)he come’_prs.sbjv > *[ˈvɛnja] > *[ˈvɛ(ɲ)ɲa] > OIt. vegna, ORo. vie) they diverge unexpectedly in the third person plural, only Italian showing the same allomorph that occurs in the first person singular and present subjunctive (e.g., ueniunt > *[ˈvɛnjont] > *[ˈvɛ(ɲ)ɲon] > OIt. vegnono but (old) Ro. vin(u)). Elson attributes the Romanian pattern to the influence of contact with middle Bulgarian, in which a pattern of allomorphy similar to that found in Romanian had emerged historically in the present tense of the verb, for independent reasons. Elson’s main argument for attributing the Romanian development to contact with middle Bulgarian is that other “eastern” Romance varieties do show the historically expected pattern in the third person plural present indicative. Since only Romanian deviates from that pattern, Elson argues, and since one thing that distinguishes Romanian from other eastern Romance varieties is precisely contact with middle Bulgarian, and since middle Bulgarian displays a similar morphological pattern to that found in Romanian, then we may conclude (it is argued) that what we see in Romanian is due to contact with middle Bulgarian. Except that we cannot conclude that at all: we cannot do so because the assumption that “eastern” Romance languages other than Romanian all display the expected pattern of allomorphy is plain wrong, and clearly vitiated by “standard language bias”. Elson (2017, 869) invokes a dichotomy between the “Romanian distribution and its deviation from the expected distribution, attested in Italian”. In this he treats “Italian” as a proxy for the whole of “eastern” Romance (cf. also Elson, 2017, 846 n. 5), paying no heed to the evidence of other Italo-Romance dialects. Yet what these latter actually reveal (cf. Maiden 2021) is that the particular morphological pattern found in Romanian, lacking the predicted palatalized allomorph in the third person plural present indicative, is that generally found in Italo-Romance as well (e.g., Scanno, Abruzzo, 1sg [ˈvjeŋgə] but 1sg [ˈvjenənə]). In fact, it is Italian (with some other central Italian dialects) which “deviates” from the general eastern Romance pattern: other Italo-Romance varieties – and indeed Romance varieties outside “eastern” Romance – pattern with Romanian. Since Italo-Romance dialects have had no contact with middle Bulgarian and, overwhelmingly, no significant contact with any other Slavonic variety, yet still turn out to share the relevant morphological pattern with Romanian, the inference that the behaviour of Romanian must reflect contact with Bulgarian simply collapses. But what made that thesis possible in the first place was the illegitimate elevation of standard Italian to the status of representative of an entire branch of the Romance languages.

I have tried in this concluding section to sketch, albeit briefly, some more Romance examples – in addition to the case of fieri in the Romanian verb ‘to be’ – of the dangers for historical linguists of according unwarranted importance to structures found in standard languages. The “moral” of all the examples here considered is the same: that historical analysis of some phenomenon in any language should never lose sight of the comparative evidence provided by its cognate, and perhaps far more obscure and less prestigious, congeners.