RRL, LXIV, 1, p. 3–28, Bucureşti, 2019
CONJUNCTION/COMPLEMENTIZER DE
ION GIURGEA1
Abstract: The origin of the conjunction/complementizer de constitutes an open problem, as none of the etymologies proposed until now is satisfactory.
Following Sava (2012), I distinguish consecutive/coordinative de (de1) from temporal/conditional de (de2), as these uses point to different origins: de1 must come from a deictic adverb meaning ‘thus, and then’, indicating temporal sequence and result, whereas de2 originates in a temporal subordinator. I argue that the most likely etymon of de1 is Latin inde, although other possibilities (dein, unde or dē+hīc/hāc) cannot be completely excluded. For de2 I propose Lat. unde. Relative de (de3) represents a secondary development of de1 or de2. I argue that ORo. înde and the MoRo. inde (a regional term from Transylvania) do not continue Lat. inde, but represent two independent developments of unde.
Keywords: etymology, pseudo-coordination, Romanian, historical linguistics.
1. INTRODUCTION
The conjunction/complementizer de constitutes an open problem in Romanian linguistics, both regarding its syntactic analysis and its origin. Regarding its synchronic analysis, the first problem concerns even its category, as can be seen from the label
‘conjunction/complementizer’. First, some terminological clarifications are in order, due to the differences between traditional grammar and modern grammars rooted in the generative tradition with respect to the categorization of clause-relating elements. I will avoid terms that are potentially ambiguous such as ‘conjunction’2. I will use the term ‘coordinator’ for the traditional ‘coordinating conjunction’ and the generative ‘conjunction’, the term
‘subordinator’ as a cover-term for traditional ‘subordinating conjunctions’, and the term
‘complementizer’ for subordinators which cannot be assigned to other categories such as adverbs or prepositions (thus, I do not reserve this term for subordinators introducing complement clauses, but I also include subordinators introducing adjunct and relative clauses).
* I am very grateful to Donca Steriade and Dana Zamfir for their valuable comments on preliminary versions of this article.
1 The “Iorgu Iordan – Alexandru Rosetti” Institute of Linguistics of the Romanian Academy, Bucharest, [email protected].
2 The syntactic behavior of what are traditionally called ‘coordinating’ and ‘subordinating conjunctions’ is so different that their inclusion in a single category ‘conjunction’ is disputable and has been rejected in generative grammars. I will use the term ‘conjunction’ only in the meaning
‘conjunction of constituents’, ‘constituent formed by coordination’.
Whereas a subordinator status is undisputable, there are contexts in which de behaves as a coordinator, the clearest situation being when the clause which follows de has the imperative mood (the imperative cannot be embedded in Romanian):
(1) Mergi de-mi adu dosarul!
go.IMPV.2SG de-me.DAT bring.IMPV.2SG file-the
‘Go bring me the file!’ (Caragiale, O. I, 160, in DLR s.v. de)
Even in one of its uses where it is generally described as a complementizer, namely, when it introduces what looks like the complements of various classes of verbs, it shows a paradoxical behavior in that it heads clauses with independent modality even with verbs normally selecting irrealis complements (see (2)) and with aspectual verbs (see (3)), which lack even independent tense (Landau 1999). Thus, ‘x wanted’+de+p in (2) is interpreted as
‘x wanted p and thus p happened’; in (3), ‘x stopped de not-p’ is interpreted as ‘x stopped p (and therefore p was no longer the case)’.
(2) Dumnezeu a vrut de nici n-a adiat vântul măcar!
God has wanted de not-even not-has breezed-softly wind-the at-least (non-standard Modern Ro.: Frăteşti, Ilfov county, in Vulpe 1980:97)
‘It was God’s wish that there wasn’t even the slightest breath of wind.’
(3) Numa atuncea se oprea de nu plângea only then REFL stop.IMPF.3SG de not cry.IMPF.3SG
‘It was only then that (s)he would stop crying’
(non-standard Modern Ro.: Burzuc, Bihor county, in Vulpe 1980:115)
The independent modality and tense, manifested in the use of the indicative in all its tenses, with their normal deictic interpretation (i.e., referring to the utterance-time), indicate a coordinator status. However, in these examples the clause introduced by de also provides somehow the content of the internal argument of the verb in the first clause – note that the verbs vrea ‘want’ and se opri ‘stop’ require an object (an internal argument), and in (2)-(3) the material in the de-clause provides the content of the object. Thus, if we adopt a coordinator analysis, we should say that de triggers somehow deletion of the complement of the first member of the coordination, under identity with the relevant part of the second member (for (3), all the verbal projections except for those introducing modality, tense and negation; for (2), we should only exclude modality and probably tense, which is in principle dependent after bouletic verbs):
(3)´ [se-opri-IND.IMPF [plâng-]] [de [nu plâng-IND-IMPF [tplâng]]]
(2)´ [vrea-IND.PF [nu adia-]] [de [nu-adia-IND-PF [tadia]]]
This type of asymmetric coordination, in which the second member provides the content of a missing complement of the verb in the first member, has been described in the linguistic literature under the name of pseudo-coordination (see Croitor 2017 for pseudo- coordination in Romanian and de Vos 2005 for a detailed treatment of pseudo-coordination in English and Afrikaans). A pseudo-coordination analysis for this type of de has been proposed by Sava (2012). A clear example of pseudo-coordination from modern standard Romanian is (4)a. Besides the fact that the conjunction here is the general coordinator şi
‘and’, a piece of evidence that the second clause is not an object clause comes from interpretation: (4)a asserts the completion of the action reported by the second clause ((s)he finished writing the letter), as expected if the mood, aspect and tense (indicative perfective past) of the second clause are interpreted, as in a main coordinate clause; however, the complement clause of the verb ‘begin’ never implies completion of the started action, as can be seen from (4)b and from its English counterpart:
(4) a. S-a apucat şi a scris scrisoarea.
REFL-has started and has written letter-the
‘(S)he began to write the letter, and (s)he wrote it.’
b. S-a apucat să scrie scrisoarea.
REFL-has begun SBJV write.3 letter-the ‘(S)he began to write the letter.’
|≠ (S)he wrote it
If we use de, we get the completion interpretation in (4)a, rather than the interpretation in (4)b (where (s)he may or may not have finished writing), which shows that de behaves as a coordinator in a pseudo-coordination construction, rather than as a subordinator:
(5) S-a apucat de a scris scrisoarea.
REFL-has started de has written letter-the
‘(S)he began to write the letter, and (s)he wrote it.’
An analysis of pseudo-coordination constructions is a complex task I will not undertake in this article, which is dedicated to the history of de. For our purposes, it suffices to recognize that what have been treated as complement clauses in de are instances of pseudo-coordination (with the exception of de să clauses and indirect interrogatives, see sections 2.1 and 2.2 below). I will call this use of de ‘pseudo-completive’.
Before getting to the historical part, I will briefly present the attested uses of de, which are much more developed in old Romanian and contemporary non-standard varieties than in modern standard Romanian (section 2). As I am interested in the origin of de, I will provide examples from Old Romanian, without considering its evolution from Old to Modern Romanian. I will then proceed to the etymological issue, the main goal of this paper (section 3). In both the descriptive and the diachronic part, I heavily rely on Sava’s (2012) dissertation, which summarizes the various etymological proposals and provides a detailed picture of the uses of de in Old Romanian.
2. USES OF DE
Sava (2012), following Roques (1907), identifies two major classes of uses, which may rely on different etyma. As these classes roughly correlate with the normal position of the clause introduced by de, we may refer to them as ‘postposed’ and ‘preposed’ de;
however, as the type for which the normal position is preposed also developed some postposed uses, it is safer to use the labels de1 and de2. Moreover, as the historical relation of the relative complementizer de with these two types is not fully clear, I will use the label
de3 for this type of de. Here is a tableau of these types, exemplified under (7) and treated in detail in the rest of the section:
Uses Position
(6) de1 : result, coordination + result, pseudo-purpose (purpose + result), postposed pseudo-completive
(+subjunctive): purpose, completive (subordinator)
de2: conditional, temporal preposed /
indirect interrogative postposed
de3: relative postposed
(7) a. de1: Mă doare de-mi vine să urlu me hurts de-me.DAT comes SBJV scream.1SG ‘It hurts so badly that I feel like screaming’
b. de2: De-aş fi ştiut, aş fi venit de-would.1SGPRF known would.1SG PRF come ‘If I had known, I would have come’
c. de3: ăla de miroase puternic (non-standard Modern Ro.: Vulpe 1980: 136) that de smells strongly
‘the one that smells strongly’
2.1 De1
De1 includes the pseudo-completive use briefly described in the introduction. This use is naturally derived from the function of introducing result clauses (the consecutive use), which is attested since the oldest texts and is preserved in the modern standard language (where the pseudo-completive use has disappeared)3:
(8) După aceea eu amu fost sărac, de n-am avut cu ce mă plăti after that I have been poor de not-have.1SG had with what me pay.INF
‘Then, I was poor, so that I could not pay for myself’ (DÎ, II, Gorj, 1563–1564) Sometimes the consequence relation between the two connected clauses is inferred by considering the event described in the second clause as the goal of the action described in the first clause. In such cases de is claimed to have a final (purposive) use (cf. Zafiu 2016), but we should notice that the use of the indicative correlates with a realis interpretation, which is not found in bona fide purpose clauses:
(9) ci ne-a<u> venit de ne-au fost domni (DÎ, XVIII, Târgovişte, 1599) which us.DAT-have come de us.DAT-have been rulers
‘who came here so to be our sovereigns (and so they were)’
(10) o deade la meşter de o legă (DÎ, LIX, Galaţi, 1570–1571) it.ACC gave.3SG to craftsman de it.ACC bound.3SG
‘She gave it (= a psalter book) to a craftsman to bind it (and he bound it)’
3 The complementizer use was still attested in non-standard regional varieties in the second half of the 20th century (Vulpe 1980: 97, 115, 119), with a reduced frequency.
Given this interpretation, de can be analyzed as a coordinator, with a richer meaning than ‘and’, including, besides logical conjunction, the purpose or consequence relation (Densusianu 1938, Rosetti 1986, Dragoş 1995, Gheţie et al. 1997, Pană Dindelegan 2016 include de both under coordinators and subordinators; for the difficulty to decide between a coordinative and subordinative status in the case of the purpose use, see Avram 1960, Dragoş 1995: 116, 120, Gheţie et al. 1997: 361, Nedelcu, 2008: 646, Zafiu 2016). An argument in favor of a coordinator status is the existence of examples in which the second clause has an imperative:
(11) pasă de te pocăiaşte (CC2 1581: 5, apud Zafiu 2016) go.IMPV.2SG de REFL.2SG repent.IMPV.2SG
‘go and repent’
We may call this use ‘pseudo-purpose’, as it involves, like the pseudo-completive use, a richer meaning (with an intended result component) added to a coordinating element.
Note that bona fide subordinate purpose clauses never imply the achievement of the intended result.
When the matrix verb itself, even if it does not require a clausal complement, includes a meaning of command (e.g. trimite ‘send’, se sfătui ‘take counsel with’), de may be analyzed either as a pseudo-purpose coordinator or as pseudo-completive coordinator.
A clear pseudo-completive coordinator status can be assumed for cases when the matrix verb requires a complement. This type is rare in the earliest attested stage (the 16th century; see Gheorghe & Mîrzea Vasile 2013, Hill & Alboiu 2016) and increases in the 17th-18th centuries, after which it decays, disappearing from the standard language (but surviving in non-standard varieties, see Vulpe 1980):
(12) au poruncitu de au făcut un sicreiu (Ureche, 178, apud Hill & Alboiu 2017:174) has ordered de have made a coffin
‘He ordered (them) to make a coffin (and so they did)’
The verbs and expressions taking de + indicative belong to two types4: (i) verbs/expressions for which the occurrence of the event described in the second clause is a result of the event in the matrix clause, either a necessary result (apuca, începe ‘begin’, avea obicei, obişnui ‘use to’, se deprinde ‘get accustomed to’, face ‘make’ (causative), cuteza ‘dare’, se apuca ‘start’, găsi vreme ‘find time to’, isprăvi ‘finish’, păzi ‘take care to’, se învrednici ‘succeed’) or an attempted/envisaged result, whose fulfillment is asserted by the de-clause (pune, porunci, învăţa ‘command’, îndemna ‘urge, advise’, stărui ‘insist’, zice ‘say’ with the meaning ‘command’, vrea ‘want’, se ispiti ‘attempt’, nevoi ‘strive to, attempt’, sta ‘insist to’, avea voie ‘be allowed to’, lăsa ‘allow’, cere ‘ask’, ajuta ‘help’); (ii) impersonal and raising verbs which imply the occurrence of the event in the complement clause: se întâmpla, se prileji, nimeri ‘happen’, fi ‘be’ with the meaning ‘happen’, ajunge, sosi, cădea ‘to come to’5:
4 This inventory is based on Drăganu (1923), Sava (2012) and Hill & Alboiu (2016).
5 To these verbs, the modern Romanian dialectal data reported in Vulpe (1980) add one attestation of plăcea ‘like’:
(13) aşea lăcomindŭ la <a> altuia, sosescŭ de pierdŭ şi al său thus coveting to (-the)-other.GEN turn.3PL de lose.3PL also the REFL.POSS
‘Thus coveting someone else’s belongings, they end up losing also their own’
(Costin, L. 89, apud Hill & Alboiu 2017: 176)
As shown in (3) above, cessation verbs take negation in the de-clause (beside se opri, I found an example of this type with se lăsa ‘cease’, cited by Drăganu 1923: 272).
The limitation of de + indicative to these verbs and the realis interpretation clearly indicate that the origin of this construction is the result clause6: first, it is precisely for these verbs that V(p) can have ‘p happened’ as a consequence. Secondly, result clauses, unlike final clauses, refer to actual results of the main clause – more precisely, the propositions they introduce are evaluated at the same possible worlds as the matrix clause (if the proposition in the matrix clause is claimed to be true in the real world, so is the proposition in the result clause, and so on), which correlates in Romanian with the use of the indicative mood.
This type of de-clauses contrasts with genuine complement clauses, represented by infinitives and subjunctives, by the realis interpretation.
Hill & Alboiu (2016) claim that the realis interpretation was not obligatory before the 18th century, but the examples they provide are not convincing. I discuss their alleged counterexamples in detail in an Appendix.
The pseudo-completive use with verbs of the type ‘happen, come to’ explains, via ellipsis, the collocation cum de ‘how was/is it possible that’, still current in modern standard Romanian:
(14) Cum de n-am ştiut?
how de not-have.1SG known
‘How could I not know?’
De1 may also combine with the subjunctive (introduced by the marker să), in final and complement clauses, in which case the realis interpretation is suspended, the de-clause being synonymous with infinitives and subjunctives. In such cases, de is a subordinator. In modern Romanian, de + subjunctive can be found in result clauses, the subjunctive indicating a potential result:
(15) Copilul... se făcuse un băiat de să nu te înduri de el child-the REFL had-become a boy de SBJV not REFL bear.2SG of him
‘The child had become such a boy that one could not bear [to separate from] him’
(Ispirescu, Legende 141, apud Drăganu 1923: 267)
(i) Îi mai plăcea de bea 3SG.DAT more like.IMPV.3SG de drink.IMPV.3SG
‘(S)he liked to drink sometimes / (S)he also liked to drink’
In this case, plăcea ‘like’ + an activity complement has a reading in which it implies the performance of this activity on several occasions. The verb in the de-clause is indeed an imperfect with a habitual reading.
6 For the realis interpretation of de + indicative, see Drăganu (1923), Frâncu (2000), Repina (2006), Sava (2012).
De1 is also found in Aromanian (with the form di) and Meglenoromanian (with the forms di or dă), with the same uses (result, see (16)a-b, purpose + achieved result, see (16)c, temporal sequence or purpose + achieved result, see (17)a–b, pseudo-completive – i.e. achieved result –, see (16)d; the examples below, except (17)b, are taken from Drăganu 1923: 260–272):
(16) a. Se aspar căpârle di-ń fug (Ar., Papahagi, Basme, 26) REFL get-frightened.3PL goats-the de-me.DAT run.3PL
‘My goats get frightened and (so that) they run away’
b. Plăndzi di ţĭ-aspardzi ocl´il´ (Ar., ibid. 3) weep.2SG de you.DAT break.2SG eyes-the
‘You’re weeping so much/so hard that you’re breaking your eyes’
c. Si duţi di lu află (Ar., Jahresbericht II, 50, 29) REFL goesde him meets
‘She goes to meet him (and she meets him)’
d. Cum putuşĭ di’nviţaşĭ ahtare băteare muşată cu fluiara?
how could(PRF).2SG de learned.2SG such play beautiful with pipe-the ‘How were you able to learn to play the pipe so beautifully?’
(Ar., Obedenaru, Texte macedo-române... de la Cruşova, II, 46)
(17) a. Si toarnă di lę trei tăciuni prinşi (Megl., Papahagi, Megleno-Români REFL turns-back de takes three embers glowing I, 101)
‘She turns back and takes three glowing embers’
b. ia-li di li speală (Megl., Capidan 1935, s.v. di) take.IMPV.2SG-them de them wash.IMPV.2SG
‘Take them and wash them / Go and wash them!’
Examples of Aromanian di in pseudo-coordination with aspectual verbs can be found in ALAR:
(18) Acăţă di-alǵaşti perlu. (ALAR I, map 16, point 19) began.3SG de- whitens hair-the
‘The hair began to turn white’
In the ALAR map 16, for the verb ‘turn gray (about the hair)’, I found this construction (‘began/begins’ + de + ‘whitens’) in 8 cartographic points (3, 4, 6, 8, 9, 19, 25, 35).
2.2 De2
De2 is mostly used as a conditional complementizer:
(19) de veţi priimi mine, priimi-veţi cela ce m-au tremes if will.2PL receive me receive-will.2PL the-one that me-has sent
‘If you receive Me, you will receive the One who has sent Me’ (Coresi, Tetr. 138v) As it introduces a conditional, it mostly precedes the main verb, by which it differs from de1 which is always postposed. Like in various other languages, including the rest of
Romance, the conditional complementizer is also used to introduce indirect polar interrogatives. In this use, de2 normally follows the matrix verb:
(20) Lasă să vedem de va veni Ilie să mântuiască el let.IMPV.2SG SBJV see.1PL if will.3SG come Elias SBJV save.3.SBJV him
‘Let’s see whether Elias will come to save him’ (Coresi, Tetr. 64v) The causal and concessive use of de, noticed in some studies (see Drăganu 1923, Gheorghe & Mîrzea-Vasile 2013, Zafiu 2016) are contextual values of the conditional one (the concessive is usually accompanied by an ‘even’-focal particle: şi, măcar).
Although it is the most frequent conditional complementizer in Old Romanian, de is certainly not the oldest conditional marker, which is se/să (< Lat. si). Indeed, in the most archaic texts, such as Codicele Voroneţean, it has been noticed that de2 has a temporal value and the normal conditional complementizer is se (Roques 1907, Zafiu 2016)7: (21) Deaci, de venreră cătră elu, dzise cătr-înşii (Cod. Vor. 9v)
then when came.3PL towards him said.3SG towards them
‘Then, when/as they came towards him, he said to them: (...)’
Greek: ὡς δὲ παρεγένοντο πρὸς αὐτόν, εἶπεν αὐτοῖς (Acts 20.18) as but came.3PL to him told.3SG them
Slavonic: i jakože priidošę kŭ nemu, reče kŭ němŭ and as came.3PL to him said.3SG to them
As shown by Drăganu (1923:257), instances of a temporal use (‘after’) can also be found in folkloric poetry, as an archaism.
The evolution from a temporal subordinator to a conditional one was repeated later in Romanian – ORo. deaca/deca/dacă ‘after’ > MoRo. dacă ‘if’.
De2, as well as de3, appear to be absent in Aromanian and Meglenoromanian8.
7 The temporal value is hard to distinguish from a conditional one when the tense is future, as in the following example:
(i) vreame de voiu dobândi, chiema-te-voiu (Cod. Vor. 32v: Acts 24.25) time de will.1SG obtain call-you.ACC-will.1SG
In this example, the Slavonic and Greek versions have a participle as V1 + main verb as V2 (Slavonic vremę že polǫčivŭ prizovǫ tę, Greek καιρὸν δὲ μεταλαβὼν μετακαλέσομαί σε), so de does not translate an explicit conditional.
For the sequence de + se there is one example with a conditional use (58r), one with a polar indirect interrogative use (22v, see (ii)) and one with an optative use in a main clause (80v), which may be related to the conditional (cf. modern Ro. de-aş fi ‘if I were’ etc.). As the conditional element here is se, it is not clear whether these uses reflect the existence of a conditional meaning of de2. (ii) dzise cătră cela ce sta, sutaşului, Pavelu, de se “omu cela rrimleanu,
said.3SG to the-one who stood.3SG centurion-the.DAT Paul de se man-the that Roman fără osându binre easte voao a-l bate?” (Cod. Vor. 22v: Acts 22.25) without condemnation good is you.DAT to-him beat
‘Paul said to the one standing by, the centurion, whether “it is right for you to beat the Roman man, without having him be [judged and] convicted” ’
8 Drăganu (1923) claims that the temporal meaning ‘since’ can be found in Aromanian, but the single example he offers is not convincing: adžun fui ş-nu ń-deditŭ măncu; şi sătos fui di nu ń-deditŭ tr s-biau (Codex Dimonie in Jahresbericht IV, 1a); the original biblical text (Matthew 25.42)
2.3 De3
De3, called a ‘relative pronoun’ in traditional studies (see GLR I: 161, Vasilescu 2008: 282, DLR, Sava 2012, Gheorghe 2016), should probably be analyzed as a relative complementizer, as it is uninflected and cannot combine with prepositions (in Romanian, like in the rest of Romance, prepositions must be fronted together with wh-phrases, they cannot be stranded like in English). It is found since the earliest texts (see (22)) and survives in regional varieties to these days, but has never become more frequent than the wh-relativizers (ce, care).
(22) Cinre e omul de va vrea viaţă (...) ? (Psalt. Hur. [c.1500] 28r, Ps. 33.13) who is man-the de will.3SG want life
‘Who’s the man who will desire life (...) ?’
3. THE ETYMOLOGY OF DE
3.1 De1
As we have seen in the previous section, the basic meaning which can be identified for de1 is that of result, as already noticed in Meyer-Lübke (189: §560). As de1 + indicative is still a coordinator, even in its pseudo-purpose and pseudo-completive uses, it is reasonable to assume that de1 was initially a coordinator indicating result (Dimand 1904, apud Sava 2012), paraphrasable by ‘thus, therefore’, or just temporal succession – ‘and then’. The fact that the pseudo-completive use is secondary can be seen along the historical development of Romanian: in the oldest attested stage (the 16th century), the pseudo- completive use is rare compared to the pure result use9; the pseudo-completive use increases in the 17th and 18th centuries. Analyzing the occurrences of de in DÎ, I found only 3 clear pseudo-completive uses (with pune ‘order’, se tocmi ‘agree’, avea voie ‘be allowed’) and 5 on the borderline between final+result and pseudo-completive (with trimite
‘send’, se băga ‘let oneself get involved in, undertake’, da ‘give’), compared to 25 instances of de1 that are not pseudo-completive (14 result, 4 result or pure coordinator, 7 pseudo-purpose, i.e., purpose+result); moreover, de să occurs only once, in a purpose clause. In the most archaic texts, the manuscripts with rhotacism, Hill & Alboiu (2016) report that the (pseudo-)complementizer10 use is absent. In Matthew’s Gospel from Coresi’s Tetraevanghel (1560-1561), I found no example in which de1 + indicative is used to translate a complement clause. I used for comparison Makarije’s Četvoroblagověstie, a Slavonic translation of the gospels printed in 1512 in Wallachia, which is likely to be the
has ‘and’ in the place of Arom. di, not ‘since’ as Drăganu interprets it: ‘I was hungry and you didn’t give me to eat, I was thirsty and you didn’t give me to drink’ (ἐπείνασα γάρ, καὶ οὐκ ἐδώκατέ μοι φαγεῖν, ἐδίψησα, καὶ οὐκ ἐποτίσατέ με). In the other sources I consulted, I found no example of de2 or de3 in the South-Danubian dialects.
9 For the reduced frequency of complementizer de in the 16th century, see Gheorghe & Mîrzea Vasile (2013), Hill & Alboiu (2016), Gheorghe (2016).
10 Hill & Alboiu call ‘complementizer de’ what we refer to as ‘pseudo-completive de’.
original of the Romanian version printed by Coresi (see Dimitrescu’s preface to Coresi, Tetr.), and, for a better understanding of the structures, the Greek text which constitutes the remote original of the Slavonic text (I used a philological edition of the Greek New Testament). Most occurrences of de1 translate the sequence anterior participle – main verb (‘anterior participle’ refers to the Slavonic perfect active participle, translating a Greek aorist participle), following the schema in (23), exemplified in (24)–(25):
(23) V1.PTCP V2.IND/IMPV → (Ro.) V1.IND/IMPV de V2.IND/IMPV
(24) Ro.: şi mearse de se spânzură (Coresi, Tetr. 62r: Matthew 27.5) and went.3SG de REFL hanged.3SG
Slavonic: i šédĭ udávi-sę and go.PTCP.PRF.MSG.NOM hanged.3SG-REFL Greek: καὶ ἀπελθὼν ἀπήγξατο and leave.PTCP.AOR.MSG.NOM hanged.3SG.MID
(25) Ro.: Scoală de ia coconul şi muma lui şi get-up.IMPV.2SG de take.IMPV.2SG child-the and mother-the his and fugi în Eghipet (Coresi, Tetr. 3r: Matthew 2.13) run.IMPV.2SG in Egypt
Slavonic: Vŭstávĭ poími otróčę i m‹a›t‹e›rĭ ego get-up.PTCP.PRF.MSG.NOM take. IMPV.2SG child.ACC and mother.ACC his Greek: ἐγερθεὶς παράλαβε τὸ παιδίον καὶ τὴν μητέρα αὐτοῦ get-up.PTCP.AOR.MSG.NOM take. IMPV.2SG the child and the mother his I found 22 examples of this type, most of them (18) with motion verbs in the first clause (‘go’, ‘get up’, ‘sit down’, ‘come’, ‘leave’, ‘descend’, ‘fall down’); the other 3 verbs are
‘take’ (+object), ‘untie’ and ‘send’.
In one of the examples, de might be interpreted as pseudo-completive (see (26);
the verb in the first clause, lăsa, means ‘leave’ but also ‘let, allow’), but this is most likely due to an error which has occurred in the printing process, because in the corresponding Slavonic text as well as in its Greek source, the second verb is singular, not plural, and has the same subject as the first one, which is an anterior participle, so that the actual meaning was not ‘he let them go’ but rather ‘leaving them, he went away’. It is known that the persons who printed the book were not the same as the translators, see Dimitrescu’s preface to Coresi, Tetr.; we may thus imagine that some person involved in the printing process, having this pseudo-completive de in his grammar, inadvertently substituted the singular which must have occurred in the manuscript Romanian translation with the plural expected here if the de-clause had furnished the content of a missing argument of the verb in the first clause, as is characteristic for the pseudo-completive use:
(26) Ro.: Şi lăsă ei de se duseră (Coresi, Tetr. 34r: Matthew 16.4) and left/let.3SG them de REFL went.3PL
Slavonic: i ostávlĭ ix otíde
and leave.PTCP.PRF.MSG.NOM them went-away.3SG Greek:καὶ καταλιπὼν αὐτοὺς ἀπῆλθεν and leave.PTCP.AOR.MSG.NOM them went-away.3SG
In 7 examples, de translates a coordination structure – 2 with the connector ‘and’
and 5 paratactic (in 4 of them, both verbs are imperative; in the fifth case, the second verb is imperative). 5 out of these 7 examples involve motion verbs as V1 (‘go out’, ‘get up’,
‘go’, ‘come’). In all cases the events described in the two clauses are successive.
In one example, de renders a V1.PART V2 sequence where V1 is a simultaneous (‘present’) participle, but was erroneously understood as anterior (the text uses the perfective past in both clauses, which yields a succession interpretation).
I also found 2 examples where V1 is a simultaneous (‘present’) participle (correctly translated using imperfective verb forms) and one example where V2 is a simultaneous participle.
Other uses of de belong to de2 (19 conditional clauses, 1 indirect interrogative, 1 reason clause, translating a postposed anterior participle11) and de3 (4 adnominal relatives, translating adnominal participles, 1 free relative translating a definite participle). In one case de occurs after the copula ‘be’ and might be interpreted as a pseudo-completive de1, with the copula interpreted as ‘happen’, but since the original text has a participle as V2, we may also think of relative de:
(27) Ro.: Era amu de avea agonisit mult (Coresi, Tetr. 41r: Matthew 19.22) was.3SG now de had.3SG possession much
Slavonic: bě́ bo iměǫ stežanïa mnóga was.3SG for having possessions many Greek: ἦν γὰρ ἔχων κτήματα πολλά.
was.3SG for having possessions many ‘for he had many possessions’
De să (i.e. de + subjunctive) is rare (5 occurrences: 2 in purpose clauses, 1 in a complement clause, 2 in main clauses with a directive force, where the Slavonic text has da+indicative).
We can conclude that de1 in the language of Coresi’s Tetraevanghel is essentially a coordinator (see especially the fact that in 4 examples the clause introduced by de has an imperative verb), but is not equivalent with ‘and’; it is used to explicitly indicate succession, and probably also (intended) result (at least in some cases). Therefore, it is mostly used to render asymmetric relations between two verbs, in which one of the verbs is a participle in the original12. However, we cannot be sure that pseudo-completive de1 did
11 The example is
(i) Ro.: Greşii de vândui sânge nevinovat (Coresi, Tetr. 62r = Matthew 27.4) did-wrong.1SG de sold.1SG blood innocent
Slavonic: sŭgrě́šixŭ prědávĭ krŭvĭ nepovínǫ did-wrong.1SG deliver.PCTP.PRF.MSG.NOM blood innocent.FSG.ACC
Greek: ἥμαρτον παραδοὺς αἷμα ἀθῷον did-wrong.1SG deliver.PTCP.AOR.MSG.NOM blood innocent ‘I have sinned having betrayed innocent blood’
12 This use is not a peculiarity of Coresi’s Tetraevanghel, but can be found in other 16th century texts. Here is an example from Codicele Voroneţean (ms. dated 1560-1580, representing a copy of an earlier translation):
(i) Ro.: venre întru pâlcu de spuse lu Pavelu (Cod. Vor. 26r: Acts 23, 16) came.3SG in troop de said.3SG the.DAT Paul
not exist in the language of the translators of Coresi’s Tetraevanghel: as the pseudo- completive use is also found in the South-Danubian dialects (see section 2.1 above), it is possible that it represents a common Romanian development and its absence in the translated texts is due to the fact that it had no counterpart in the original text (such a productive use of a pseudo-coordination construction is crosslinguistically exceptional; it is clearly absent from New Testament Greek). The possibility that the pseudo-completive use of de1 represents a more recent development which took place independently in Dacoromanian and the South Danubian dialects cannot be excluded either (the fact that this use is rare in the 16th century original texts, as I could confirm by examining the occurrences of de in DÎ, supports this hypothesis).
To conclude, the origin of de1 must be sought in an adverb showing result or temporal sequence, equivalent to ‘thus’, ‘and then’. The fact that the result meaning must be old is demonstrated by its presence in Aromanian and Meglenoromanian. As de1 is, thus, of what is called “common Romanian” age and since such adverbs are rarely borrowed, it is the most likely that de1 continues a Latin word or, at least, is a Romanian creation based on Latin words. The Turkish origin suggested in Meyer-Lübke (1899: §560) must certainly be rejected. A substrate origin, also suggested by Meyer-Lübke (loc. cit.), based on the comparison with Alb. dhe ‘and’, and adopted by Tiktin (TDRG I s.v. de), can of course never be completely excluded (as the pre-Roman languages of the Balkans are unknown), but the comparison with Albanian dhe is problematic: first, this word means ‘and, also’, lacking the temporal sequence or result additional meaning which characterize de1. Secondly, the Albanian interdental fricative dh- in word-initial position does not correspond to Romanian d- (see Rosetti 1986: 242-243, Brâncuş 1983: 13). Moreover, Çabej (1986: I 151) argues that dhe is a shortened form of edhe ‘and, also’. Orël (1998: 85) proposes that edhe comes from Proto-albanian *e (= Albanian e ‘and’, < Indo-Eur. *ōd) + *de (< Indo- Eur. *dō, cf. Slavonic da) with d>dh in intervocalic position, but there is no independent evidence, in Albanian, for the existence of this *de. In any case, as I have already said, the probability for a word of this type (an adverb meaning ‘and then, thus’) to be borrowed is small.
However, no convincing Latin etymon has yet been proposed as yet. The preposition de ‘from’, later ‘of’, which has become an important functional preposition in Romanian, as well as in the rest of Romance, has been proposed as the etymon of de1,2,3 by Drăganu (1923)13 – a proposal adopted in many studies: Scriban (1939), Procopovici (1948), Iordan (1954), Ciorănescu (1966), ILR II: 292, DLR, Sava (2012), Gheorghe (2013) – but, as Meyer-Lübke (1899: §560) already noticed, Lat. de could never have become an adverb or coordinative connector. Indeed, the transformation of a very frequently used monosyllabic preposition into an adverb has no parallel in the Romance domain. The general tendency goes in the opposite direction: the prepositions of Latin, as well as other Indo-European languages, generally originate in particles that could occur without a complement, functioning thus as (spatial) adverbs; after they start taking obligatory complements, they
Slavonic: vŭšĭdŭ vŭ plŭkŭ, sŭkaza Pavŭlu enter.PTCP.PRF.NOM.MSG in troop said.3SG Paul.DAT
Greek: εἰσελθὼν εἰς τὴν παρεμβολὴν ἀπήγγειλε τῷ Παύλῳ enter.PTCP.AOR.NOM.MSG in the troop announced.3SG the.DAT Paul.DAT 13 Previously, Philippide (1894:51–52) had proposed Lat. de as an etymon, but only for de2 in its conditional use. Drăganu (1923) extended this proposal to all uses of de (de1,2,3).
never return to the stage where the complement was optional. New adverbs come from prepositional phrases based on nouns or other adverbs (see împotrivă ‘against’ < în potrivă, afară ‘outside’ < Lat. ad foras, înainte ‘before’ < în + ainte < a + *inte < Lat. (ab)ante, apoi ‘then’ < Lat. ad post, îndărăt ‘back’ < în + *dărăt < Lat. de retro, etc.; cf. Väänänen 1967:99 for the tendency of reinforcing adverbs in the evolution from Latin to Romance).
Moreover, the semantic development from an ablative preposition to an element which indicates result or succession is highly unlikely14. Manoliu’s (2006) proposal that de represents an extension to finite forms of the preposition de used with infinitives cannot account for the coordinating use of de and for the characteristic realis meaning of de + indicative (therefore, the verbs or nouns that take de + infinitive are different from those that take de + indicative, see Sava 2012:118); moreover, as Jordan (2009:42-43) shows, de + infinitive is very rare in the oldest texts – it is absent from Coresi’s Lucrul apostolesc and his 1577 bilingual Psalter and is attested only once in Codicele Voroneţean – which indicates that it is more recent than de1.15
Schuchardt’s hypothesis (in Literaturblatt für germanische und romanische Philologie, 1892, 204, apud Meyer-Lübke 1899: §560) that de1,2,3 reflects a mixture of Latin de with South-Slavic da is likewise unacceptable because the preposition de and the Slavic complementizer da have a totally different distribution – da is a complementizer which introduces finite clauses, whereas the preposition de cannot directly combine with a finite verbal form. Moreover, Slavic da corresponds to the Romanian subjunctive să marker, being found in irrealis clauses16, whereas the peculiarity of de, as I have shown, is the realis interpretation (unless it combines with the subjunctive marker să).
There are in fact several possible Latin candidates for an adverb/sentence connector de meaning ‘and then; thus, therefore’, but they involve certain irregular phonetic developments, which may explain why they have not been considered in previous studies.
We should look for an ablative deictic adverb – ‘from there/here, from now/then’, which can indicate both result and succession, similar to Old Ro. de aci ‘from here’ which yielded Modern Romanian deci ‘thus, so’. The most suitable candidate, given the meaning and
14 Such an evolution is proposed by Drăganu (1923), who derives de2 as well as de1 from the temporal use of Latin de ‘since; after’. But, notwithstanding the difficulty of the change of distribution from preposition to (adverbial) complementizer taking finite clauses, the meanings of the preposition de can at most explain de2, assuming that its initial meaning was temporal anteriority. As de1 indicates an event occurring after the event described in the matrix, possibly being its result, its meaning is rather the opposite of the meaning of de2 and the preposition de.
15 Densusianu (1938: 410) admits the possibility of an extension from de + infinitive only in the case of de + subjunctive.
16 Here are examples from Coresi, Tetr.:
(i) vru amu Irod să ceară coconul să-l piarză voao (Coresi Tetr. 3r) wanted now Herod SBJV search.3.SBJV child-the SBJV-him destroy.3.SBJV you.DAT
Slavonic: xoštetŭ bo Irodŭ iskati otročęte, da pogubitŭ je (Matthew 2.13) wants for Herod search child.GEN so-that destroy.PFV.3SG him
‘For Herod wants to look for the child, in order to destroy him’
(ii) nu vrea mine să fiu împărat (Coresi, Tetr. 164v) not want.IMPF.3 me.ACC SBJV be.1.SBJV emperor
Slavonic: ne chotęvŭšęję mĭně da cěsarĭ bimĭ bylŭ (Luke 19.27) not want.PTCP.PRF.ACC.PL me.ACC that emperor be.OPT.1SG been
‘who didn’t want me to be emperor’
frequency of use it had in Latin and the extent to which it was preserved in the Romance languages, is inde, which had both spatial and temporal meanings – ‘from there, thence, thenceforward, thereafter, thereupon, then’. As shown in REW 4368, this word is pan- Romance (besides Romanian înde, which I will discuss below, REW cites Old It. inde, nde, It. ne, indi, Veronese de, Old Logudorese nde, Engadinese, Friulian in, end, Fr. en, Occ. en, ne, Catalan ne, Old Sp., Old Asturian, Old Portuguese ende). Interestingly, this word has become a clitic in many varieties, undergoing the higher degree of phonetic reduction specific to clitics (see It. ne, Fr., Occ., Cat. en). For Romanian, we may obtain de by a similar irregular phonetic reduction, from the form *nde/n̥de expected for common Romanian if inde had become unstressed (it has been argued that the sequences unstressed i- and a- + tautosyllabic nasal evolved to syllabic nasals in common Romanian, and the î- which we find in modern Dacoromanian represents an epenthetic vowel, see Puşcariu 1928:
780, Petrovici 1930: 70–71, Densusianu 1938: 32, Avram 2012: 82–88; for the interpretation of the word-initial ä used in the earliest texts for the present-day unstressed în-/îm- as a syllabic nasal, see Avram 1964: 125-126, 1990: 65, 76–8017). A form nde is also found in Old It., and its reduction to de is attested in Veronese.
For the loss of the initial nasal of the forms *nde/n̥de, due to the phonetic erosion characteristic of function words, we have parallels in Aromanian: Lat. intrō > (î)ntru > tru, tu; Lat. intrā > ntră > tră.
A potential problem for this etymology is the fact that, according to the Dictionary of the Romanian Academy (DA), Romanian would have another form traced back to inde:
the adverb inde ‘where’, used in Transylvania, with an archaic form înde meaning ‘when, as’ (Cod. Vor. 31v,5). DA explains the unexpected semantic evolution via a contamination with unde ‘where’. This explanation is not very convincing. Scriban (1939) prefers to see înde, inde as mere variants of unde; TDRG treats inde as a variant of unde and has a special article for the archaic înde, for which it suggests as etymon the same unde18. I think TDRG’s solution is correct. The form inde may owe its i- to the influence of dinde ‘from where’ (mentioned by DA s.v. inde), which might have appeared spontaneously from de- unde in fast speech – cf. the contraction pănde < pă unde. The form înde, which appears in a few of the earliest texts, should indeed be separated from inde ‘where’, because it has a different meaning and distribution – it is a temporal and conditional subordinator, never attested as an interrogative word (the following list exhausts the attestations of înde I’ve been able to find in the Old Romanian texts, using indices and the digitalized corpus of the
“Iorgu Iordan – Al. Rosetti” Institute of Linguistics):
(a) in Cod. Vor. 31v (Acts 24, 20) it corresponds to an absolute anterior participle in the original (înde stătuiu în gloată – stavšŭ mi vŭ sŭnĭmišti ‘as I stood before the council’);
(b) in Psalt. Hur. 41v (Psalm 48, 18) it corresponds to Slavonic vŭnjegda ‘when’
(Coresi’s bilingual Psalter has no subordinator here);
(c) in Psalt. Hur. 45v (Psalm 54, 13) and 69v (Psalm 80, 14) it corresponds to Slavonic ašte, Greek εἰ ‘if’ (Coresi’s bilingual Psalter has conditional să/se in both examples);
17 See Avram (1962, 1968, 1986) for arguments in favor of the view that the support vowel î- is absent from the underlying phonological representation even in modern Romanian (important evidence comes from the behavior of this vowel in sandhi).
18 Unfortunately, the new editions of TDRG (Wiesbaden, Otto Harrassowitz, 1988 and 2003) abandon this etymology (without even citing it!), adopting DA’s etymology instead.
in Cod. Bratul19 it corresponds to:
(d) a preposed anterior participle20 in 51 (înde veni – prišedše ‘having come’), 121 line 17 (e-nde auziră – slišavše že ‘having heard’), 128 (e-nde trecură întânia paze şi a doao – prišedša že prĭvǫę straže i vtorǫę ‘having passed the first guard and the second’), 130 line 8 (e-nde deşchiseră – otvrĭzše že ‘when/after they opened [the gate]’), 155 (E-nde binevestiră ciătăţiei aceiia – blagověstvovavša že gradu tomu ‘after they preached the gospel to that city’), 158 (e-nde veniră întru Ierusalim – prišedšę že vŭ Ier‹u›s‹a›l‹i›mĭ ‘and after they came to Jerusalem’), 190 (e-nde auziră di-mvierea morţilor – slišavše že ot vŭskrěšenia mr‹ŭ›tvi‹xŭ› ‘when they heard about the ressurection of the dead’), 199 (e-nde auziră elu Achila şi Prischila – slišavše ‹že› ego Akilla i Priskilla ‘Aquila and Priscilla, having heard him’), 201 (e-nde auziră – slišavše že ‘having heard’), 208 (e-nde auziră elu – slišavše že i ‘when they heard and...’ – here i ‘and’ seems to have been erroneously translated with elu ‘him, it’), and 230 (înde văzură elu – viděvše ego ‘having seen him’);
(e) a preposed absolute anterior participle in 129 (e-nde bătu Petru poarta curţiei – tlŭknŭvšu že Petru vrata dvoru ‘when Peter knocked at the gate of the yard’), 130 (e-nde fu ziuă – d‹ĭ›nĭ že byvšu ‘as soon as it was day’), 178 (e înde zi fu – d‹ĭ›ni že byvšu ‘as soon as it was day’), 223 (E-nde rădică-se Chiprulu şi lăsemu elu de-a stânga – vŭzniku že Kipru i ostavlĭšǫ ego ot šuǫ ‘after Cyprus appeared and (we) outstripped it on the left hand’), 227 (e-nde fumu noi întru Ierusalim – bivše že nam‹ĭ› vĭ Ier‹o›s‹o›l‹u›mě ‘when we arrived to Jerusalem’), 234 (E multe înde fură pări ‘as there were many dissensions’ – mnodzě že bivši raspri ‘as there was much dissension’), 236 (e-nde adunară-se cicea – sŭšedšem že se zde ‘when they had gathered here’), 258 (e-nde acestuia fu ‘after this happened to him’ – semu že bivšu ‘after this happened’) and 261 (e-nde adunarăse ei – sŭšedše že im‹ŭ› ‘when they had gathered’);
(f) a postposed absolute anterior participle in 121 line 6 (înde pomeniiu ‘when/as I recalled’ – poměnǫše že ‘recalling’; here Coresi L. has când pomeniiu, and the Greek original, Acts 11.16, has a main clause – ἐμνήσθην δὲ τοῦ ρήματος τοῦ Κυρίου ‘I remembered Lord’s word’) and 321 (înde se deşertă moartea – upraznivše že se sĭmrtĭ
‘death being abolished’);
(g) conditional ašte ‘if’ in 341 (înde amu Dumnezeu îngerii ceia ce greşiră nu-i cruţă – ašte bo B‹og›ŭ a‹rxa›gg‹e›li sŭgrěšivšeę ne poštǫdě ‘if God did not spare the angels who sinned’; here Cod. Vor. has se ‘if’ and Coresi L. has când ‘when’); in 362 and 363, -nde is preceded by de ‘from, of’ yielding the form di-nde, corresponding to Slavonic ašte (362: di-nde aşa îndrăgi noi Dumnezeu – ašte sice vŭzljubi nas‹ŭ› (bogŭ ‘God’ omitted):
‘since God loved us so’; 363: cine di-nde spune-va că... – iže ašte ispověstĭ jako... ‘whoever will declare/confess that...’; the co-occurrence of a relative and a conditional marker is probably due to the conditional marker ajn in the Greek original: John’s Epistle I, 4.15 ὃς ἂν ὁμολογήσῃ ὅτι...);
(h) in five cases, the original has no subordinate clause: 148 (Acts 13.49-50: e-nde purtase cuvântul Domnului prin toate laturile, e ovreaii întăriră curatele mueri... – pronašaaše že se slovo g‹ospod›ne po vŭsei strani, Iudei že naustišǫ častiviǫ ženi... ‘Lord’s word was spreading over all countries, but the Jews stirred up the honorable women ...’),
19 I am grateful to my colleague Dana Zamfir for having informed me of some of the attestations of înde in Cod. Bratul and of the attestations of de1,2 in the same text, and for having checked the indices of a number of old Romanian texts for the presence of înde.
20 Under ‘participles’ I include the invariable forms called ‘gerundives’ in Vaillant (1958).
149 (e-nde răsipiră-se mulţimea cetăţiei – razdělišę že sę množestvomĭ grada ‘and the crowd of the city was divided’; here the instrumental form množestvomĭ may have suggested an absolute participle construction; the Greek original, Acts 14.4, has coordination: ἐσχίσθη δὲ τὸ πλῆθος τῆς πόλεως, καὶ οἱ μὲν ἦσαν σὺν τοῖς Ἰουδαίοις, οἱ δὲ σὺν τοῖς ἀποστόλοις ‘and the people of the city were divided, some were for the Jews, some were for the apostles’), 169 (e-nde sosi întru Dervie şi în Listră – prispě že vŭ Dervii i Lystrǫ ‘(when) he came to Derbe and Lystra’), 406 (e-nde împarte-se şi muiarea şi fata ce nu mărită-să – razděli se ž‹e›na i d‹ě›va[a] noposagšia bo ‘woman and unmarried girl are separated’; there follows a break in the text); in 405 line 2, the relation between the two clauses is interpreted as conditional in the original: e-nde legi-te muieriei, nu ceare dezlegarea – privezaeši že se ženě, ne išti razdrěšenia ‘you are bound to a woman: don’t seek to be freed’ (= 1 Corinthians 7.27 δέδεσαι γυναικί, μὴ ζήτει λύσιν; Coresi, L. has conditional să here);
(i) in one case, 226, it looks as though the Romanian main verb corresponds to a postposed participle in the original, and the verb of the (î)nde-clause corresponds to the main verb (E-nde nu supuindu-se elu tăcum, zisemu ‘After we had stopped talking, as he did not obey, we said...’ – ne povinuǫštu že se mu umlĭčaxom‹ŭ› rekšu ‘as he did not obey, we stopped talking, saying...’ = Acts 21.14 μὴ πειθομένου δὲ αὐτοῦ ἡσυχάσαμεν εἰπόντες);
another possible analysis is that the translator erroneously combined two ways of rendering the anterior participle, first using înde and then the gerund (supuindu-se ‘obeying’);
(j) In Iorga’s Apostle (Galatians 3.25, p. 476 in Gafton’s edition of Cod. Bratul) it corresponds to a preposed absolute anterior participle: înde veni credinţa – prišedšię věry
‘when faith has come’, Greek ἐλθούσης δὲ τῆς πίστεως).
As can be seen from all these examples, Ro. înde does not correspond semantically to Lat. inde. It probably represents the result of a reduction undergone by unde ‘where’
when used as a temporal subordinator, a context in which we can assume that it had become unstressed. As discussed above, word-initial în- (written ä or än) in 16th century Romanian probably represents a syllabic nasal. So, it may be assumed that unstressed unde was reduced to nde (see the frequent form e-nde ‘and when/if’), realized as n̥de (conventionally transliterated as înde) when there was no vocalic support for the initial n-. This account is supported by the fact that unde itself is also used as a temporal subordinator in Cod. Bratul (the text where the most occurrences of înde are found) and can be used to translate an anterior participle, exactly like înde:
(28) Ro. E unde auzi Anania cuvintele aceastea (Cod. Bratul 47) and where heard.3SG Anania words-the these
Slavonic: slyšavše že Anania slovesa sia hear.PTCP.MSG.NOM PTCL Ananias words-the these
‘When Ananias heard these words...’
Note also that where Cod. Bratul 362 has the form di-nde ‘since, if’ (see under (g) above), Coresi, L. has d-unde, lit. ‘from where’). Both de unde ‘from where’ and unde were sometimes used to introduce conditional clauses in ORo. (Densusianu 1938: 284, 289).
As in section 3.2 I will propose that de2 itself comes from Lat. unde, we may consider this înde, -nde an older form of de2. Notice however that de2 itself is present in Cod. Bratul, although it is not very frequent. It translates Slav. ašte ‘if’ (64 in an indirect
interrogative; 427, 432, 435, 455 in conditionals) and direct interrogatives in 40, 98 and 242. As de2 is only found in conditionals and uses derived from the conditional, whereas înde is predominantly temporal, these forms cannot be considered variants. I conclude that înde reflects a later reduction of unde.
In sum, it is unlikely that Lat. inde is continued by Rom. inde/înde, because of their totally different meaning and distribution. The most likely etymon of both these forms is unde ‘where’.
It is generally agreed that Latin inde has been preserved in some compound forms:
Old Romanian decinde ‘on the other side’ (Megl. diţindea), formed with the preposition de and the deictic particle ecce, Aromanian dinde (< de + inde), didinde < (de + dinde) ‘on the other side’, Old Romanian tutindere, tutindene ‘everywhere, all over’ < tot(um) + inde + re (> Mo. Ro. pretutindeni ‘everywhere’). As in these contexts inde was stressed, the phonetic evolution has been different.
Another possible Latin etymon for de1 is dein, a shortened form of deinde = dē + inde. This form had predominantly a temporal meaning in Latin, ‘thereafter, then, next’, which is suitable for de1. In Romance languages, the form deinde was preserved in Old Venetian dende, Occ. den, Sp., Old Port. dende (REW 2525), as well as in Aromanian (see above). The evolution dein > de1, however, is not fully regular either: final -n in monosyllables is preserved in in > în (the monosyllabic pronunciation of -ei- in deinde and dein is indicated by Latin poetry); a preserved -n < -m might be found in can ‘rather’ < Lat.
quam (can > Modern Ro. cam, a more recent form, see DELR II; however, Lat. quam has a second result in Romanian, namely ca ‘than, as’, in which the final nasal was lost); in cum
‘with’, the final nasal was preserved as -n before vowels at an unattested stage of the language, which explains the form cunus(ul) ‘with him’ < * con ipso) (+ the article –l). Besides, we should assume that the diphthong -ei was reduced to -e: dein > *dei > de.
An adverb meaning ‘from here/there’ could also have appeared from the preposition de + a deictic adverb; possible combinations which may have yielded de are dē hīc and dē hāc; hīc ‘here’ has been preserved in Romanian in eccum/*accu- hīc > aci; from dē hīc, we should assume a form *dei (with diphthong) > de; hāc ‘(by) this way, on this side’ (a perlative adverb) is preserved as a locative deitic adverb in many Romance languages, in composition with eccum/ecce (see REW 3965), and possibly also in the particle -a of Rom.
acesta (‘this one’)21; from dē hāc, we should assume a form *dea (with diphthong) > (in unstressed position) de.
Given that all the possible etymons involve special reduction processes, I prefer those based on inde – inde and dein – because they not only had a frequent use in Latin, but are also preserved in other Romance languages. Given that the loss of an initial nasal syllable is also to be assumed for de2, as I will show in the next section (3.2), and that inde is pan-Romance, I consider inde as the most likely etymon. The transformation of a deictic adverb into a sentence connector and then a coordinator is a natural process, also found with Rom. şi ‘and’ < Lat. sīc ‘so, like that’.
The pseudo-completive use of a coordinator element is a phenomenon found in other Balkan languages (see Scr. te, Bg. ta, Modern Greek kai, Alb. e, cf. Sandfeld 1904, 1930, Skok 1973). Sandfeld (1930) gives examples of the verb ‘order, command’ followed by a
21 Cf. Drăganu (1936/1938: 263), Rosetti (1986: 373). The preservation of hāc in Arom. aoa (< ad hāc), proposed by Candrea & Densusianu (1914), is unlikely, because it does not explain the -o-.
coordinator, with the interpretation ‘x ordered p and therefore p happened’, in Serbo- Croatian (with te ‘and’), Albanian (with e, the regular word for ‘and’) and Greek (with kai, the regular word for ‘and’), besides Ro. de. He also cites examples with ‘begin’, ‘decide’
and ‘allow’ for Modern Greek kai, with ‘can’ for Albanian e, Modern Greek kai, Macedonian i and Aromanian di, with ‘want’ for Modern Greek kai, Alb. e, Arom. de.
Weigand (1904, Krit. Jahresbericht über die Fortschritte der rom. Philologie, apud Sandfeld 1930) noticed that Romanian is more similar to Serbian and Western Bulgarian, and differs from Modern Greek and Albanian, in that (i) the coordinators used in these environments are different from the normal word for ‘and’ – cf. Ro. de vs. şi, Scr. te vs. i, Bulg. ta vs. i, and (ii) Ro. de and Scr. te can also introduce relative clauses. Therefore, unlike Sandfeld, who proposes that all these phenomena have their origin in Greek, I consider that, if contact is involved in the special evolution of de, we should restrict it to languages which have a similar ‘result coordinator’ distinct from the unmarked coordinator
‘and’, i.e. Serbo-Croatian and Bulgarian. Whether for these languages we may assume the influence of Ro. de, or the other way around, is a question which cannot be answered based on the data I dispose of. In any case, the initial meaning I reconstruct for Romanian, based on Latin inde, is compatible with an internal development. As for the Slavic forms, they are likewise derived from a deictic base (the demonstrative root to- ‘that’) and can be assumed to have undergone the same development as Ro. de1: according to Vaillant (1977:§§ 1468- 1469 and 1958:§324), Scr. te is related to Old Slavonic tě ‘then’ (compared to Lith. taĩ
‘thus, so’), and the initial meaning of Bg. ta was ‘thus, then’ (compared to Lith. tuõ ‘for that, immediately’; BER also compares Sanskrit tāt ‘therefore’, indicating as etymon an Indo-European ablative form of *to-); Scr. te is assigned the basic meaning ‘and then’ by RJHSJ, vol. 74: 138.
3.2 de2
Most researchers did not separate de1 from de2. However, since the ancestor of de1
can be reconstructed as an adverb used as a sentence connector indicating result and succession, whereas de2 was initially a temporal subordinator indicating anteriority (‘after’
or ‘since’, see section 2), it is hard to imagine a common source for these items, given the categorial difference (adverb vs. subordinator) and above all the significant semantic difference (in ‘p de1 q’, q is a result of p or follows p, whereas in ‘de2 q p’, q is anterior to p or is a condition for p). Therefore, following Roques (1907) and Sava (2012), I consider it likely that these items have different origins.
In the case of de2, the semantic difference with respect to the preposition de is no longer a problem, as the preposition de can have the temporal meaning ‘since’ and also a causal meaning. Therefore it is no wonder that Drăganu (1923: 256) proposed that de as a subordinator first appeared with the temporal meaning ‘from the moment that, since’, from which all the other uses subsequently emerged.
Drăganu’s hypothesis is untenable for a syntactic reason: in Romance languages, in Latin, as well as in other Indo-European languages, prepositions cannot directly combine with a phrase headed by a finite verb22. Therefore, in order to combine a preposition with a
22 In Romanian, an exception is până ‘until’ (Old Romanian also ‘as long as’), which can take clauses with or without complementizers (până (ce) a plecat ‘until (that) has left’); note however that this preposition differs from de in that it never combines directly with nominals, but only with
