RRL, LXIII, 4, p. 325–340, 2018
OF SECOND PERSON SINGULAR PRESENT VERB-FORMS.
THE TYPE (TU) VII, (TU) RĂMÂI
MARTIN MAIDEN
1Abstract. This study seeks to understand, from a comparative-historical perspective, the origins of Romanian second person singular present-tense verb-forms of the type vii, rămâi, alternating with the root-allomorph vin-, rămân-. It is argued that neither the historical emergence of this type, nor its subsequent resistance to analogical levelling, can be satisfactorily explained in ways traditionally proposed in the Romanian historical linguistic literature. Appeals to phonological conditioning as the explanation of the alternant’s emergence, and to avoidance of homonymic clash as the explanation of its subsequent persistence, seem untenable. Rather, at work in both cases is the analogical influence exercised over verbs having root-final nasals by a pattern of alternation characteristic of verbs having root-final dentals. This abstract and essentially morphological analogical conditioning is argued to operate even when the dental alternation might still be considered to be phonologically conditioned.
Keywords: Romanian; verb; iotacization; morphology; analogy; morphomic structure; homonymic clash
1. ‘IOTACIZATION’ IN ROMANIAN
‘Iotacization’ (in Romanian, iotacizare) occupies a prominent place in Daco- Romance historical phonology and morphology2. It properly denotes historical processes of palatalization and/or affrication of consonants triggered by immediately following yod or, under certain circumstances, by immediately following [i]. However, the term ‘iotacized’
(iotacizat) is also used to describe the resultant morphological alternations. Iotacization alternations, affecting root-final consonants, permeate Daco-Romance verb morphology, and have the following phonological manifestations:
1 Research Centre for Romance Linguistics, University of Oxford, e-mail: martin.maiden@mod- langs.ox.ac.uk.
2 See, for example: Morariu (1927; 1929); Puşcariu (1937a: 70–73; 1937b: 257–259); Ivănescu (1948=2012); Pop (1952); Iancu (1965; 1966); Vasiliu (1968); Sala (1976: 119–136; 225–232);
Saramandu (1992); Calotă (1986: 276–278); Donovetsky (2011); Maiden (2011).
non-iotacized root-final consonant iotacized root-final consonant t [t] ~ ţ [ʦ]
d [d] ~ z [z] (older [ʣ]) s [s] ~ ş [ʃ]
n [n] ~ i [˜]3 (r [r] ~ i [˜])
Examples of these alternations may be observed in in Table (1), which shows the present, and third person subjunctive,4 forms of a number of verbs. The phenomenon principally and most extensively affects non-first conjugation verbs: in the first conjugation, only the second person singular present is ever affected, and root-final consonants [n] or [r] are (almost) never affected at all.5 In the non-first conjugation verbs, iotacization is much more extensively represented in older Romanian, or in modern dialects, than it is in the twenty-first century standard language. For this reason I have placed in parentheses, alongside their modern equivalents, alternants which are no longer current. The nature and history of the elimination of these iotacized alternants will be addressed later in this study.
Table (1). Effects of iotacization (underlined) in verb morphology
a. Iotacization in first conjugation verbs: cânta ‘sing’, uda ‘wet’, suna ‘sound’, ara
‘plough’, lăsa ‘let’
PRS SBJV PRS SBJV PRS SBJV PRS SBJV PRS SBJV
1SG cânt ud sun ar las
2SG cânţi uzi suni ari laşi
3SG cântă cânte udă ude sună sune ară are lase lase
1PL cântăm udăm sunăm ară lase
2PL cântaţi udaţi sunaţi araţi lăsaţi 3PL cântă cânte udă ude sună sune ară are lase lase b. Effects of iotacization (underlined) in non-first conjugation verbs: simţi ‘feel’, vedea
‘see’, ieşi ‘go out', coase ‘sew’, veni ‘come’, sări ‘jump’:
3 The pronunciation tends to be [˜] syllable- or word-finally, and [j] syllable-initially, although Romanian linguists often transcribe [˜] in both contexts. Because the difference between these variants seem genuinely allophonic, and for clarity of exposition, I have here transcribed the sound as [˜] in either context.
4 The first and second person forms of the subjunctive are, almost without exception, identical to those of the present, so need no further separate consideration.
5 There are at most three first conjugation verbs in which we may find iotacization in the first person singular present and the third person subjunctive. These are mâna ‘drive (cattle)’, amâna
‘postpone’ (e.g., 1SG.PRSmâi, amâi, 2SG.PRSmâi, amâi, 3SBJVmâie, amâie) and, in some dialects speria 'frighten' (1SG.PRSspai, 2SG.PRSspai, 3SBJVspaie). In the first two verbs we have a kind of suppletion, in which forms of the non-first conjugation verb (ră)mâne 'stay, remain', with inherited iotacization, have intruded in to their paradigms (see Morariu 1927: 305; Lombard 1954/1955: 270–273).
The position with speria is more difficult, but again the existence of a non-first-conjugation variant of this verb seems to be part of the explanation (see Morariu 1927: 288, 297; Lombard 1954/1955: 941f.).
PRS SBJV PRS SBJV
1SG simt (simţ) văd (văz)
2SG simţi vezi
3SG simte simtă (simţă) vede vadă (vază)
1PL simţim vedem
2PL simţiţi vedeţi
3PL simt simtă (simţă) văd vadă (vază)
1SG ies cos
2SG ieşi coşi
3SG iese iasă coase coasă
1PL ieşim coasem
2PL ieşiţi coaseţi
3PL ies iasă cos coasă 1SG vin (viu)
2SG vii
3SG vine vină (vie)
1PL venim
2PL veniţi 3PL vin vină (vie) 1SG sar (saiu)
2SG sari (sai)
3SG sare sară (saie)
1PL sărim
2PL săriţi 3PL sar sară (saie)
The origin and morphological distribution of the iotacized alternants is mostly uncontroversial. They are attributable to the effects of a proto-Romance yod, which produced palatalizing or affricating effects on immediately preceding consonants. Thus, the forms historically underlying 1SG.PRS simţ and 3SBJV simţă, 1SG.PRS văz and 3SBJV vază, 1SG.PRS viu and 3SBJV vie, and 1SG.PRS saiu and 3SBJV saie are, respectively, *'sentjo (< SENTIO), *'sentja (< SENTIA(N)T), *'vedjo (< UIDEO), *'vedja (< UIDEA(N)T), *'venjo (< UENIO), *'venja (< UENIA(N)T), *'saljo (< SALIO), *'salja (< SALIA(N)T). The second person singular forms cânţi, simţi, uzi, vezi, laşi, ieşi (pronounced kɨnʦʲ, simʦʲ, uzʲ, vezʲ, and usually laʃ, jeʃ) are also phonologically explicable through a Daco-Romance rule of palatalization/affrication of the relevant consonants in the historical environment of an original following [i]. This alternation unfailingly affects the second person singular of all verbs containing the relevant root-final consonants. There was also widespread analogical extension of the effects of iotacization by yod into other non-first conjugation verbs in root- final -[t], -[d], -[n], and -[r], which would not, historically, have undergone phonological iotacization because they did not originally contain yod (see, e.g., Pop 1952: 206–208).
Typical examples are vinde ‘sell’ (1SG.PRS vânz, 3SBJV vânză for expected, and modern standard, vând, vândă), scoate ‘remove’ (1SG.PRS scoţ, 3SBJV scoaţă for expected, and
modern standard, scot, scoată), pune ‘put’ (1SG.PRS pui, 3SBJV puie for expected, and modern standard, pun, pună), cere ‘ask for’ (1SG.PRS cei, 3SBJV ceie for expected, and modern standard, cer, ceară). Further, dialectal, examples may be seen in ALRII maps 1922, 1926, 1941, 1945, 2057, 2060, 2062. My concern here, however, is the nature and origin of the paradigmatic distribution of the alternations,6 rather than their lexical extensions (for a discussion of the theoretical significance of the analogical spread and retreat of these alternations through the Romanian lexicon, see, e.g., Maiden 2011).
2. THE PROBLEM OF THE SECOND PERSON SINGULAR PRESENT TYPES VII, PUI, SAI, CEI
Romanian non-first conjugation verbs whose roots end in [n] and [r] show the alternant [˜] in the second person singular present.7 In the case of root-final [n], this is true almost without exception; in the case of root-final [r], the alternant [˜] is absent in the modern standard language, but well attested in the older written language and in dialects.
No other kind of alternant is attested: we either find [˜], or we find no kind of iotacization in this context. My view is that the presence of [˜] in the second person singular present is doubly anomalous in the history of Romanian and the Daco-Romanian dialects. It is anomalous because its existence in that particular cell of the paradigm cannot be explained as an effect of sound change, and it is also anomalous because, in the face of later elimination (analogical levelling) of the iotacized alternants, as observable in Table (1), the [˜] alternant shows obstinate persistence just in the second person singular: exceptionlessly in the case of verbs in [n], very commonly for verbs in [r]. So, not only is [˜] in these cases a historical ‘intruder’, but it is also a ‘squatter’, in that it stubbornly refuses to leave when threatened with eviction. To account for these aberrant historical behaviours, I need first to explain why I dissent from two widely held views, namely that second person singular present forms such as vii, pui, sai, cei are in fact the result of sound change, and that their subsequent failure to be eliminated is motivated by avoidance of potential homonymic clash with the first person singular.
3. THE TYPES VII, PUI, SAI, CEI ARE NOT THE RESULT OF SOUND CHANGE
It has been assumed, notably in an influential paper by Pop (1952: 199, 214–217), but also, for example, by Ivănescu (1948=2012: 245, 259, 264), Iancu (1965: 285), or Saramandu (1992: 83), that iotacized second person singular forms of verbs with root-final
6 The iotacized alternant generally also appears in the gerund. The origins and history of iotacization in the gerund are discussed in detail in Maiden (2011). Since the gerund is not particularly relevant to the behaviour of the second person singular present forms, which are at issue here, I do not discuss the gerund in this study.
7 I am assuming here that [˜] is the segment that alternates with [n]. In principle, [˜] could also be analysed as the second person singular inflexional ending, in which case the lexical root has no final segment. The question is probably otiose, since a sequence of root-final [˜] followed by an inflexional ending [˜] would in any case resolve as a single [˜].
nasals (or root-final [r]) have a purely phonological origin, as direct reflexes of proto-forms
*'puni, *'vini, *'sari (< *'sali), and so forth. However, iotacized second person singular forms such as vii, ţii, pui, spui, rămâi, cei, sai do not appear explicable phonologically. The Latin etyma for these forms are UENĪS, TENĒS, PONIS, EXPONIS, REMANĒS, QUAERIS, SALĪS. It is likely that all of these endings (-ĪS, -ĒS,-IS), emerged as *-[i] in proto-Daco-Romance (for detailed defence and illustration of this view see, e.g., Maiden 1996), and the comparative evidence of modern Daco-Romance varieties points consistently to an original common second person singular ending *-[i], so that the proto-forms would have been something like *'veni, *'tjeni, *'poni, *es'poni, *re'mani, *'keri, *'sali. Now, final *[ni],
*[ri], *[li] regularly yield, respectively, [nj], [r]j, [rj], where [j] reflects the normal development of final unstressed *-[i] under most phonological circumstances. That this is the normal development is shown not only by first conjugation verbs, such as suni [sunj]
‘you sound’, but equally by nominal morphology, where the reflex of *-[i] is a plural marker: e.g., buni [bunj] ‘goodMPL’ (< *'boni), câini [kɨ˜nj] ‘dogs’ (< *'kani), oameni ['wamenj] ‘people’ (< *'omeni), surori [su'rorj] ‘sisters’ (< *so'rori), pari [parj] ‘poles’
(< *'pali), meri [merj] ‘apple trees’ (< *'meli). Indeed, non-iotacized second person singular forms of the type puni (or puri reflecting rhotacism of intervocalic [n]), showing the regular phonological development, are well attested in early Romanian texts where, in contrast, the corresponding first person singular present and third person subjunctive forms do show iotacization (cf. Sala 1976: 229).
Now [˜] actually is a phonological reflex of nasals or laterals originally followed by
*[i], but only where – unlike the foregoing examples – those nasals or laterals were originally long (Puşcariu 1937: 257f.; Vasiliu 1968: 133; Gheţie and Mareş 1974: 158, 163;
Sala 1976: 228–232; Avram 1990: 159)8. The original development is *[nni] > [ɲ], and
*[lli] > [ʎ], a phase well preserved in trans-Danubian dialects. In most Daco-Romanian varieties, both of these palatals merge as [˜]. Thus *'anni ‘years’ yields ai [a˜] in old Romanian (and in modern dialects of north-western Transylvania, Crişana, and Maramureş, and of Moldova east of the Prut). In fact ai happens to be the only clear example of this development for the nasal in Daco-Romance, but there is more evidence for the lateral: e.g.,
*'valli > văi ‘valleys’, *ka'valli > cai ‘horses’, *'kalli > căi ‘ways’, *'elli > ei ‘they’. In contrast, single *[n] and *[l] (the latter becomes [r] intervocalically, e.g., *'pali > pari) never emerge as [˜] when followed by *[i]. In short, second person singular present forms such as vii, pui, sai do not seem explicable phonologically. A second person singular present form in [˜], alternating with [n] or [r], is always of purely morphological origin.
8 The existence, in some dialects, of plurals [plu'mɨ˜] or [plə'mɨ˜] ‘lungs’ (see, e.g., NALRBanat map 95, points 37 Ferendia, and 68 Delineşti), corresponding to standard Romanian plămâni, is not evidence for a sound change *[n˜] > [˜], although it has sometimes been so interpreted (see Byck and Graur 1967: 58; Cazacu 1966: 126f.). The fact that it occurs in Banat, where [˜] is not the regular output of palatalization of [n], perhaps suggests that we may be dealing with some kind of hypercorrection, speakers equating their indigenous development, namely [ɲ], with standard Romanian [˜]. Note that in the examples given above, the singular also contains yod (e.g., plu'mɨ˜e), apparently analogically extended from the plural.
4. HOW DID THE TYPE VII, ŢII, RĂMÂI, (SAI) EMERGE?
If the iotacization in second person singulars of verbs in root-final [n] (and [r]) has no phonological origin, then where does it come from? I submit that it is no coincidence that the [˜] alternants only appear in verbs which also have, or had, the iotacized alternant in the first person singular present and the third person subjunctive. This, by the way, is why the [˜] alternant never appears in the first conjugation, for first conjugation verbs lack iotacized first person singulars present and third person subjunctives. Recall that [˜] (the outcome both of [ɲ] and of [ʎ]) has a regular historical phonological explanation in the first person singular present and third person subjunctive of such verbs, where it is the reflex of
*[nj] and *[lj]: e.g., REMANEO, REMANEA(N)T > *re'manjo, *re'manja > *re'mɨɲu, *re'mɨɲa
> rămâiu, rămâie; SALIO, SALIA(N)T > *'saljo, *'salja > *'saʎu, *'saʎa > saiu, saie. The conclusion seems to me inescapable that the [˜] found in the second person singular present of these verbs is analogically introduced from those other cells of the paradigm where there is an iotacized alternant. But why? The most plausible explanation is one that seems to have enjoyed relatively little favour among Romanian scholars, apart from Vasiliu (1968:
112) and Sala (1976: 229), namely Sextil Puşcariu's appeal to a kind of proportional analogy based on the iotacization pattern of verbs with root-final dentals. Recall that in these latter, the iotacized alternant systematically appears not only in the first person singular present and in the third person subjunctive, as a result of proto-Romance affrication before yod, but also in the second person singular present, as a result of later Daco-Romance affrication before [i]: e.g., UIDEO, UIDES, UIDEA(N)T > *'vedjo, *'vedi,
*'vedja > *'veʣu, *'vedi, *'vaʣa > văz, vezi, vază; SENTIO, SENTIS, SENTIA(N)T > *'sentjo,
*'senti, *'sentja > *'senʦu, *'senʦi, *'senʦa > simţ, simţi, simţă. On the model of these dental verbs, where the same iotacized alternant is thus shared by first person singular present, second person singular present, and third person subjunctive, verbs with the iotacized alternant [˜] in the first person singular present and third person subjunctive introduce that alternant into the second person singular present. This is what is proposed by Puşcariu (1937b: 257f.):9
Par analogie avec quelques verbes tels que eu auz (< audio) – tu auzi (audis), on a introduit dans les verbes ayant un n à la fin du radical la consonne mouillée aussi à la seconde personne: eu viń – tu vińi (au lieu de vini), eu ţiń – tu ţińi (au lieu de ţini). En dehors du dialecte istro-roumain, qui conserve aujourd'hui encore les formes tu viri (< vini), tu ţiri (ţini), tous les autres connaissent exclusivement les formes tu ţii (< ţińi), tu vii (< vińi), aroum. tine ţîńi, tine yińi, mégl. tu [ţ]ń, tu viń.
This explanation, invoking morphological analogy, seems the only available and plausible one for the second person forms. Note, however, that the type of analogy involved is relatively abstract, since it says, in effect, that just as in verbs with root-final dentals the alternant found in the first person singular present and third person subjunctive present is
9 Puşcariu only discussed verbs with nasal roots, and assumed that the analogy took place when the iotacized alternant was still [ɲ], in his transcription, ń.
also found in the second person singular present, so in verbs with root-final nasals (and root-final [r]) the alternant found in the first person singular present and third person subjunctive present will also occur in the second person singular present. In terms of phonological content, the alternants implicated are quite different: on the one hand [t] ~ [ʦ], [d] ~ [(d)z], on the other [n] ~ [˜] ([ɲ]).
5. DEIOTACIZATION
5.1. Typology and geography of deiotacization
‘Deiotacization’ is the process by which original iotacized alternants are eliminated from the verb in favour of their non-iotacized counterparts. Deiotacization does not, however, usually affect the second person singular present, and this is the central question that will concern us shortly. I focus, first, on the deiotacization of the first person singular present and third person subjunctive forms, which have been subject to the phenomenon, in differing degrees, across Daco-Romance. As the examples in Table 1 above reflect, deiotacization has gathered considerable ground over the past few centuries. Of the four segments which have iotacized alternants in the relevant cells of the paradigm, namely [t], [d], [n], and [r], the dentals are those most consistently affected by deiotacization over the widest geographical area.10 Indeed, trans-Danubian dialects almost11 never show the iotacized dentals in the first person singular present or third person subjunctive, even in verbs which must, historically, have undergone iotacization: e.g., reflexes of *'audjo ‘I hear’, *'vedjo ‘I see’, *'patjo ‘I suffer’, *'vedja ‘seeSBJV’ > Istro-Romanian 'ɑvdu, ved, Aromanian 'avdu, ved, pat, ‘veadə. Dentals are also deiotacized in a continuous area comprising western, northern, and eastern Daco-Romanian, except for southern Transylvania, Munţii Apuseni, eastern Oltenia, Muntenia, and parts of southern Dobrogea.
Deiotacization of verbs with nasal-final roots in those morphological environments is much more geographically restricted. It is general in trans-Danubian dialects (e.g., *'tenjo ‘I hold’, *'venjo ‘I come’, *'tenja ‘comeSBJV’ > Aromanian 'ʦənu, 'ʦənə, 'jinu, Megleno- Romanian ʦən, vin, Istro-Romanian 'ʦiru, 'viru), but exceptions are detectable12, and the deiotacization appears to be of relatively recent date. In Daco-Romanian, deiotacization is absent in the third person subjunctive in southern Crişana, Banat, Hunedoara, and north- west Oltenia, and for the first person singular present in two distinct areas: southern Crişana, the Banat, Hunedoara, and western Oltenia, and in eastern Maramureş, northern Moldova, and northern Dobrogea. Finally, the [r] ~ [˜] alternation is truly rare, perhaps unsurprisingly if we bear in mind that the only cases of such alternation attributable to historical sound change are those where [r] is historically *[l], there having been no
10 For the geographical extent of the deiotacizations in Daco-Romanian, see Pop (1952: 218–
220), Saramandu (1992: 84–86), Donovetsky (2011:142). Also ALRII maps 1911, 1938, 1942, 1950, 1953, 1958, 2053, 2059, 2061, 2063, 2065, and ALRSinteza maps 40, 63, 228, 232.
11 But see Morariu (1927: 309).
12 See Popovici (1914: 100) for remnants of iotacization of [n] in the first person singular present in the Istro-Romanian of Brdo (e.g., 'viɲu ‘I come’). See also Morariu (1927: 308).
iotacization of *[r]. The alternation [r] ~ [˜] is completely absent from modern standard Romanian and trans-Danubian dialects (Iancu 1965: 284n30). It is absent from reflexes of subjunctive *'dolja (< DOLEA(N)T) ‘hurt’ in central and western Oltenia, Banat, Crişana, western Transylvania, western Oaş, as well as in eastern Maramureş, eastern Transylvania, Bucovina, northern Moldova, and almost all of Muntenia and Dobrogea, and absent over an even larger area for reflexes of SALIO, SALIA(N)T ‘jump’.
5.2. Resistance of the second person singular present to deiotacization, and traditional explanations of deiotacization
There is a widely held view in Romanian historical linguistics that there is a causal link between the resistance to deiotacization of the second person singular and the motivation of deiotacization. Specifically, many Romanian linguists (e.g., Morariu 1927:
308; Ivănescu 1948=2012: 252; Pop 1952: 233; Calotă 1986: 276–278; Saramandu 1992:
84; Marin, Mărgărit, Neagoe 1998: 105) argue, or accept, that deiotacization, or at any rate deiotacization of the first person singular present, is at least partly motivated by avoidance of homonymic clash with the second person singular present indicative. In a nutshell, the idea is that in alternations such as 1SG.PRS şez(u) ‘sit’ – 2SG.PRS şezi, or 1SG.PRS rămâi(u)
‘stay’ – 2SG.PRS rămâi, the phonological distinction between the inflexional endings becomes compromised as a result of sound change, making the two verb-forms actually, or practically, indistinct. This allegedly vexing homophony is resolved by replacing the first person singular allomorph with the allomorph found elsewhere in the paradigm (şed-, rămân-), whence the situation found in the modern standard language with alternations such as 1SG.PRS şed ~ 2SG.PRS şezi, or 1SG.PRS rămân ~ 2SG.PRS rămâi.
This idea seems to me to present, however, multiple difficulties. ‘Avoidance of homonymic clash’ is, generally, too blunt a tool in explaining morphological change (cf. also Lass 1997: 355–358). If some change has the effect of differentiating expressions whose forms would otherwise be identical, then it is always tempting to appeal to disambiguation of homophony as the explanation; but the role, if any, played by disambiguation in real morphological change remains scarcely understood, and there is permanent risk of falling into the trap of ‘teleologism’, by confusing the effect with the cause. Morphological forms do not exist in a vacuum, but are uttered in circumstances in which there are multiple other contextual clues as to (say) the person and number of the subject. And if real risk of ambiguity still arises, speakers generally already possess disambiguatory strategies without having to invent new ones. In Romanian, for example, potential confusion between auz ‘I hear’ and auz(i) ‘you hear’ could readily be resolved by use of an overt subject pronoun (e.g., eu auz ‘I hear’ or tu auzi ‘you hear’). The hypothesis of avoidance of homonymic clash is dogged by another, fundamental, problem which has largely been ignored, namely that the alleged process of avoiding homonymic clash tends to create it! The cost of avoiding homophony between 1SG auz and 2SG auzi by replacing the former with aud, is that there arises a new homonymic clash with the existing third person plural present form, which is aud. In fact, it is systematically true in modern Romanian non-first conjugation verbs that the third person plural present is homophonous with the first person singular present. Speakers’ alleged strategy for avoidance of homophony between first person singular and second person singular seems, then, singularly imprudent, since they leap out of the frying pan straight into the fire. If Pop
(1952: 211, 212) observes that there is a ‘tendance d'avoir des formes bien distinctes pour les trois personnes du singulier’, or that the homophony between first person singular present văd and third person plural văd is of a kind ‘tolérée par le roumain’, he merely describes the situation, rather than providing any real explanation of it.
‘Homophony avoidance’ also predicts that deiotacization should be less liable to occur where there is additional root-allomorphy distinguishing first person singular from second person singular, than where there is not: thus the vocalic allomorphy in the Daco- Romanian verb vedea ‘see’ (1SG văz – 2SG vezi) might be expected to make this verb less liable to deiotacization in the first person singular than verbs with no such allomorphy, since there is no risk of homophony. But the geographical area of deiotacization in the first person singular present of this verb is effectively identical to that of deiotacization of auz ‘I hear’, where risk of homophony with the second person singular is clearly greater (cf.
Puşcariu 1994: map 36). Just as unexpected, if the deiotacization in the first person singular present is really a matter of distinguishing it from the second person singular present, is the diachronic behaviour of putea ‘be able’. This verb originally had, across Daco-Romanian, a first person singular present iotacized form pociu. The origin of this form is problematic (see Zamfir 2005: 419, 423f., 428), but the relevant point here is that the corresponding second person singular present form has always been poţi, so that there was no risk of homophony. In most modern Romanian dialects, excepting some northern and western varieties, pociu has nonetheless been replaced by pot, showing the root-final dental characteristic of the whole of the rest of the paradigm.
Appeal to homophony avoidance also seems blithely to ignore the fact that Romanian verb morphology is shot through with enduring syncretisms (mainly of phonological origin), without any sign of general pressures to eliminate them. Thus, distinctions of number have for centuries been systematically neutralized in the third person of all subjunctives (3SG.SBJV cânte, facă, doarmă = 3PL.SBJV cânte, facă, doarmă), and in the first conjugation present (3SG.PRS cântă = 3PL.PRS cântă); the distinction of person and number is now always neutralized between the first person singular and third person plural present in non-first conjugation verbs (1SG.PRS fac, dorm = 3PL.PRS fac, dorm); the distinction between present and imperfect is neutralized in the second person plural of first conjugation verbs (2PL.PRS cântaţi = 2PL.IPF cântaţi), and so forth. This is not to say that there are no cases which might be interpreted as disambiguation (cf. old Romanian imperfect 1SG cânta = 3SG cânta = 3PL cânta > modern 1SG cântam ~ 3SG cânta ~ 3PL cântau), but it is never clear that a disambiguatory effect reflects a disambiguatory motivation, especially if the change actually results in new homophony, as it does in the case of the new 1SG cântam which, replacing earlier cânta,becomes identical to 1PL cântam.
One consideration that might, however, lend some support to the hypothesis that deiotacization is motivated by homophony avoidance is the fact that, in dialects where iotacized forms are recessive but not extinct, there is sometimes an asymmetry such that the first person singular is deiotacized, but the third person subjunctive is not (see Maiden 2011 for a survey)13. Since the third person subjunctive never (in any relevant case) shows
13 Iotacization is also extensively preserved in the gerund, another environment in which there is no risk of ambiguity: cf. modern Romanian 1SG.PRSvăd ‘see’, 3SBJV vadă, GER văzând; 1SG.PRS
scot ‘remove’, 3SBJV scoată, GER scoţând. However, the gerund in Daco-Romanian is a major locus of aberrant, both conservative and innovative, morphological behaviour which is by no means limited to the iotacized forms. For further discussion, see Maiden (2011: 79f.).
syncretism with other forms of the verb, there might be argued to be no need for deiotacization. This asymmetry, which is principally observable for the [n] ~ [˜] alternation, may readily be seen, for example, in ALRII maps 1938/2059 or 1958/2065 (also Puşcariu 1994:334/map 49), especially in Maramureş, Bucovina, and Moldova, where the type 3SBJV rə'mɨje ‘stay’, 'vije ‘come’, often alternates with the type 1SG.PRS rə'mɨn, vin. The situation in western Romania is described as follows in Maiden (2011:77)14:
Beloţu (point 970 [NALROltenia]) has 1SG.PRS/SBJV vəd, sɨmt, sar, ʦɨn, rə'mɨn, vin vs 3SBJV 'vazə, 'sɨmtə, 'sarə, 'ʦɨje, rə'mɨje, 'vije. The fact that the [z] alternant survives in this dialect in the subjunctive form 'vazə is actually exceptional for central Oltenia, where the normal pattern is vəd 'vadə. Reflexes of SALIRE in central and western Romania do show occasional cases of elimination from the subjunctive rather than the 1SG.PRS/SBJV (e.g., 972 Mărăcine sa˜ – 'sarə). A survey of materials from the regional linguistic atlases (NALRBanat and NALRCrişana) for Banat and Crişana, as well as the data from ALRII vol. VII, strongly bears out this picture, with coherent levelling for root-final dentals, and largely also for root-final nasals and [r], but with a systematic pattern of asymmetry in cases of incoherence, such that levelling affects the 1SG present and subjunctive rather than the the third person subjunctive.
As Maiden goes on to show, there are also occasional ‘asymmetrical’ deiotacizations in the heartland of iotacization in central and southern Romania but, for example, data from Transylvania show only a very weak tendency to affect just the first person singular present, and there are plenty of cases where it is the subjunctive, not the first person singular present, that deiotacizes. To the extent that there is a tendency to deiotacize the latter, rather than the former, it can credibly be argued that this reflects not homophony avoidance, but the arguably higher frequency of use of the third person subjunctive over the first person singular present in Romanian (for discussion, see Maiden 2011: 77).
Crucially, the facts of Daco-Romanian deiotacization need to be seen in the wider Daco-Romance perspective. Asymmetrical deiotacization favouring the first person singular present over the third person subjunctive is, in fact, exceptional. It does not occur at all in trans-Danubian dialects and, although detectable in some Daco-Romanian varieties, it rarely affects dentals and is often found alongside the opposite type of asymmetry, in which the subjunctive is deiotacized. The fact that the subjunctive is usually just as subject to iotacization as the first person singular present is, indeed, inconsistent with appeal homophony avoidance, since the third person subjunctive is morphologically distinct from other verb-forms.
It is not just that deiotacization sometimes occurs where there is no risk of homophony, for it normally does. We have seen that deiotacization is general in Aromanian, yet here there is extensive retention of the distinction between first and second person singular inflexional endings, and good reason to believe that deiotacization significantly predates major neutralization of those endings. Overall, reflexes of both -[u]
and -[i] survive extensively, particularly after consonant clusters (Capidan 1932: 263–65,
14 I adapt some presentational details of the original.
285–287; Vasiliu 1968: 81). Thus, when Ivănescu (1948=2012: 254) adduces the Aromanian first person singular present form şăd ‘sit’ to support his view that loss of final [u] motivates deiotacization in Aromanian as well, he overlooks the fact that the historically underlying form was şădzu or şedzu, containing a phonological environment (the affricate [ʣ]) in which [u] would not have been deleted (cf. Capidan 1932: 445):15 this verb is, therefore, one in which one would not expect to find deiotacization. Istro- Romanian has total deiotacization of first person singular present forms, yet the inflexional marking of the second person singular, -[i], is robustly maintained in all verbs (Puşcariu 1926: 90), and -[u] is also quite often preserved in the first person singular (Puşcariu 1926:
98), so in Istro-Romanian, too, it appears that deiotacization in the first person singular has taken place in circumstances where there was no risk of homophony:, e.g., 1SG ved ‘see’
2SG 'vezi 3SG 'vɛde 1PL ve'dem 2PL ve'deʦ 3PL 'vedu.
In Daco-Romanian, the circumstances in which there is real risk of homophony are also much rarer than sometimes assumed. Deiotacization occurs in standard Romanian, yet here final -ʲ generally survives after [ʦ] and [z]: therefore, replacement of, say, 1SG râz [rɨz] ‘laugh’ ~ 2SG râzi [rɨzʲ] by 1SG râd ~ 2SG râzi takes place in the absence of true homophony. The perceptual distinction may be low, but it is still there. Dialects where the distinction between the endings of the first and second persons singular really is phonologically neutralized are indeed found over much of western and northern Romania, in regions where there is also deiotacization in the first person singular, but the match remains very far from perfect. Thus, final -j and/or first person singular -w is well preserved in parts of Crişana, Maramureş, and north-western Transylvania where deiotacization is also well established: cf., e.g., ALRII map 1950, where the type 1SG a'udw ‘hear’ ~ 2SG a'uz (point 250, Petreştii-de-Jos).16 The presence of deiotacization in the first person singular occurs in Romanian dialects in which original *-[u] persists as lip-rounding on the root- final consonant leads Ivănescu (1948=2012: 254) into the, frankly, circular argument that if deiotacization occurred in the presence of this persistent phonological distinction, that proves that the distinction was not sufficient – thereby contradicting his own claim (1948=2012: 258) that deiotacization only occurs where the distinction is phonologically neutralized. In any case, the allegedly undesirable homophony is widely tolerated in Daco- Romanian, at least in verbs showing root-final [˜] in first and second persons singular present. Thus ALRII maps 1938 and 1941 show, respectively, the type 1SG rə'mɨ˜ ‘stay’ = 2SG rə'mɨ˜ in twelve localities and 1SG spu˜ ‘say’ = 2SG spu˜ in seventeen, in Moldova, Dobrogea, and Muntenia. Finally, if avoidance of homophony were a motivation for deiotacization, we should expect verbs with root-final nasals to be just as susceptible to the phenomenon as those with root-final dentals, but the dialectal deiotacization of the former has a much smaller geographical extent than that of the latter.
What all the foregoing indicates is that deiotacization was not motivated in any significant measure by avoidance of homophony. But then what did cause it? My suggestion is simple: deiotacization is a common-or-garden example of analogical levelling, such that the alternant found in most of the paradigm, or at any rate in such
15 Similarly, 2SG.PRS şedză < *'ʃeʣi.
16 On the issue of preservation of marking of first person singular alongside deiotacization, see also Iancu (1966: 273n23).
prominent and highly frequent parts as the third person singular and plural present, ousts the minority alternant contained in the first person singular present indicative. So if we have văd ~ vadă, or rămân ~ rămână, for older văz ~ vază, and rămâi ~ rămâie, it is because, for example, the third person present forms, and the infinitive, and the imperfect, and so forth, contain [n] and [d] (3SG.PRS rămâne, 3PL.PRS rămân, INF rămâne, IPF
rămânea; 3SG.PRS vede, 3PL.PRS văd, INF vedea, IPF vedea). What we see is no different from ‘deiotacization’ elsewhere in Romance, as when old Italian 1SG.PRS.IND veggio ‘see’
~ 3SG.PRS.SBJV veggia yields modern vedo ~ veda (cf. 3SG.PRS.IND vede, 3PL.PRS.IND vedono, INF vedere, IPF.IND.3SG vedeva), or when older ‘iotacized’ Ibero-Romance forms such as those still observable in Portuguese 1SG.PRS.IND meço ‘measure’ (< Lat. METIOR) ~ 3SG.PRS.SBJV meça (cf. 3SG.PRS.IND mede, 3PL.PRS.IND medem, INF medir, IPF media) emerge in Spanish as 1SG.PRS.IND mido ~ 3SG.PRS.SBJV mida. Yet if the Romanian iotacized alternant is removed as a result of general analogical levelling from the first person singular and the third person subjunctive, the obvious question is why is it not equally removed from the second person singular present. Why do we get 1SG.PRS rămân and 3SBJV rămână, but not also 2SG.PRS **rămâni?
5.3. Why the second person singular present resists deiotacization
It is indeed puzzling that one never finds a second person singular present such as
**rămâni [rə'mɨnʲ]17 – especially given that in the second person singular present of first conjugation verbs this type is normal, and that it even occurs in some non-first conjugation verbs (e.g., cerni [ʧernʲ] ‘you sieve’). If one poses the same question with regard to verbs with root-final dentals, however, the answer seems patent. The reason why we always get vezi [vezʲ], simţi [simʦʲ], auzi [a'uzʲ], trimiţi [tri'miʦʲ], never **[vedʲ], **[simtʲ], **[a'udʲ],
**[tri'mitʲ], is that the latter forms would violate the constraint that word-final [dʲ] and [tʲ] are impossible18. Wherever we find a root-final dental, it alternates with [ʦ] / [z] before -[ʲ]. This is true of all first conjugation verbs, even though in every other part of the paradigm these verbs display the dental (e.g., 1SG.PRS cânt ‘sing’, 2SG.PRS cânţi, 3SG.PRScântă; 1SG.PRS ud ‘wet’, 2SG.PRS uzi, 3SG.PRS udă). It is also true of all nouns and adjectives (e.g., MSG ud ‘wet’ FSG udă
MPL uzi FPL ude; FSG coadă ‘tail’ FPL cozi; MSG lat ‘wide’ FSG lată MPL laţi FPL late; MSG frate
‘brother’ MPL fraţi), and of all neologisms (e.g., MSG byte ‘byte’ MPL byţi).
It seems to me that the only plausible explanation for retention of the iotacized alternant in the second person singular of verbs in root-final [n] is that it is reflects the analogical influence of the pattern of alternation typical of the dental class: deiotacization is blocked in verbs with root-final dentals for apparently phonological reasons but, because this happens in the second person singular of the dental verbs, deiotacization is also, analogically, blocked in the second person singular present of verbs in root-final [n] (and often for root-final [r]). Significantly, this is not the first time in the history of Romanian that the alternation [n] ~ [˜] in the second person singular has shown itself to be somehow
17 Except, for example, in dialects of western Romania where *-[ni] and *-[nj] merge phonologically.
18 The exception is final [ʃtʲ] (and the extremely rare [ʒdʲ]).
'under the control' of the alternation pattern found in dentals, for we have seen that the former initially emerged on the analogy of the latter. The two historical mechanisms I have in mind may be represented schematically as in Tables (2) and (3). In Table (2), the first phase, we have a proportional analogy such that the pattern of identity found in dental verbs between the root-allmorph of the first person singular present and the third person subjunctive of verbs, on the one hand, and that of the second person singular present, on the other, supplies the root allomorph of the second person singular present in verbs with root-final nasals.
Table 2. Analogical extension of the iotacized alternant into the second person singular present 1SG.PRS 2SG.PRS 3SG.PRS 3SBJV 1SG.PRS 2SG.PRS 3SG.PRS 3SBJV
au(d)zu au(d)zi aude au(d)ză
simţu simţi simte simţă
↓
viu *vinʲ vine vie → viu vii vine vie rămâiu *ră'mɨnʲ rămâne rămâie → rămâiu rămâi rămâne rămâie saiu *sarʲ sare saie → saiu sai sare saie
Later, the model of verbs in root-final dentals, where deiotacization has taken place, but has been blocked in the second person singular present, also regulates deiotacization of verbs with root-final nasals. This is illustrated in Table 3.
Table 3. Analogical blocking of deiotacization in the second person singular present Original iotacization Deiotacization, blocked in 2SG.PRS
1SG.PRS 2SG.PRS 3SG.PRS 3SBJV 1SG.PRS 2SG.PRS 3SG.PRS 3SBJV
auz auzi aude auză → aud auzi aude audă simţ simţi simte simţă → simt simţi simte simtă
↓ ↓ ↓
viu vii vine vie → vin vii vine vină rămâiu rămâi rămâne rămâie → rămân rămâi rămâne rămână saiu sai sare saie → sar (sai) sare sară
The notion that the alternation pattern characteristic of dentals ‘leads the way’ for the nasals is further borne out by the geographical distribution of the two types of alternation.
As shown earlier, levelling for dentals in the first person singular present and third person subjunctive has a notably broader geographical extent than that for levelling of the [n] ~ [˜]
alternation in the same environments. The dominance of the pattern associated with dentals over that associated with nasals probably also has a numerical motivation. Quite simply, there are very many more verbs with root-final dentals than there are with root-final nasals, at least in the non-first conjugation verbs that are at issue. In fact, the [n] ~ [˜] alternation is restricted at most to five verbs19, and to verbs historically or synchronically derived from
19 Spune ‘say’ (< EXPONERE) stands in a historically derived relationship to pune ‘put’
(< PONERE), but its semantic distance from pune justifies counting it as separate.
them. I give examples here in the third person singular present: spune ‘says’; pune ‘puts’;
(ră)mâne ‘stays’; vine ‘comes’; ţine ‘holds’. In contrast, verbs in root-final dentals are far more numerous, even without counting their derived forms: e.g., pierde ‘loses’, vinde
‘sells’, rade ‘shaves’, tunde ‘shears’, cade ‘falls’, prinde ‘catches’, roade ‘gnaws’, şade
‘sits’, vede ‘sees’, râde ‘laughs’, aude ‘hears’, crede ‘believes’, închide ‘closes’, decide
‘decides’, ucide ‘kills’; bate ‘beats’, trimite ‘sends’, poate ‘can’, simte ‘feels’, pute ‘stinks’, minte ‘lies’, scoate ‘removes’, împarte ‘shares’, înghite ‘swallows’, ascute ‘sharpens’.
6. CONCLUSIONS
I have argued that the pattern of ‘iotacized’ root-allomorphy in the second person singular present, characteristic of verbs with root-final dentals has, not once but twice in the history of Daco-Romance, exercised a powerful analogical influence on other types of alternation associated with iotacization. This influence is notably abstract, in that it is not a matter of analogical replication of the concrete phonological alternations [t] ~ [ʦ], [d] ~ [z]
(although this, of course, happens), but of replication of the historically accidental pattern of distinctive root-identity between first person singular present, third person subjunctive, and second person singular present and, later, of the pattern of alternation caused by the survival of a differentiated alternant just in the second person singular present.
Both of these morphological phenomena might be viewed as ‘morphomic’ in the sense that they are autonomously morphological in nature (see especially Maiden 2011).
They are recurrent and robust parts of the architecture of Romanian morphology, but it is synchronically impossible to claim that they are triggered by some non-morphological factor. The set of cells ‘first person singular present + third person subjunctive + second person singular present’ is irreducible to any morphosyntactic natural class. The first time that the ‘dental’ class exercises its morphological influence on the ‘nasal’ class there is a fairly clear structural parallel in the relevant patterns of alternation: the nasals, like the dentals, show a distinctive alternant in the first person singular and third person subjunctive, and since that same alternant appears in the second person singular present of the dental class, so that pattern of sameness is analogically introduced into the second person singular of the nasal class. When deiotacization occurs, the existing similarities of patterning between the dental class and the nasal class act in such a way that the process of elimination of the iotacized alternants in the dental class, characteristically blocked in the second person singular present, displays parallel blockage in the process of deiotacization of nasals. The course of the diachronic levelling of iotacized dentals is ‘mimicked’ by that affecting nasals.
It might be thought that phonological conditioning could still be invoked to explain some of what we observe. After all, the reason that the second person singular present form of verbs in root-final dentals becomes identical to the allomorph of the first person singular and third person is precisely affrication before original final -*[i] (> -[j]), and the reason why deiotacization is later blocked in that same cell of the paradigm is the resulting phonological constraint in Romanian that words cannot end in **-[tj] or **-[dj].20 Yet we
20 I do not intend here to investigate in detail the phonological status of the constraint against dental + [j]. The fact is that the potential for such a sequence only arises at the boundary between lexical root and inflectional ending. With the arguable exception of marţi ‘Tuesday’ (but cf. the definite form martea), all instances of final -[j] in combination with dental roots are analysable as
have seen that the iotacization, and later non-deiotacization, of the second person singular present in verbs with root-final nasals, are not explicable phonologically. In brief, however much phonological factors may have determined, and might continue to determine in synchrony, the iotacization of dentals, the resultant alternations are also plainly morphologized, in that only a morphological analogy can explain the projection of the alternation patterns into the nasal verbs. In fact, the nasal verbs are not the only source of evidence that iotacized alternant in the second person singular present of verbs in root-final dentals is morphologized independently of any phonological conditioning. In some Daco- Romanian dialects the alternation between [t] and second person singular present [ʦ] has been analogically introduced into verbs with root-internal [t], where the dental is not in fact adjacent to the putative conditioning environment. This type can be seen from ALRII maps 1891/1892 (first person singular, second person singular, and third person present tense forms of a pieptăna ‘to comb’) and 1887 (first person singular and second person singular of a vătăma ‘to wound’), illustrated in Table 4.
Table 4. Analogical extension of iotacization to root-internal dentals
1SG 2SG 3SG
551 Pipirig 'ceptən 'cepʦənj 'captənə 682 Somova 'kjeptən 'kjepʦənj 'kjaptənə 316 Sânnicolaiul-Român 'vatəmw 'vaʦəmj
987 Topraisar 'vatəm 'vaʦemj
Here, too, the iotacized alternant has clearly become morphologized, regardless of whether its original phonological conditioning environment persists. Speakers have again assigned a purely morphological analysis to alternations which, taken in isolation, might appear to invite purely phonological explanations.
ABBREVIATIONS
ALRII = Petrovici (1956- ) ALRSinteza = Saramandu (2005-) NALRBanat = Beltechi et al. (1980-) NALRCrişana = Stan and Uriţescu (2000-)
REFERENCES
Avram, A., 1990, Nazalitatea şi rotacismul în limba română, Bucureşti, Editura Academiei Române.
Beltechi, E., I. Faicuc, N. Mocanu, P. Neiescu, 1980-, Noul atlas lingvistic român pe regiuni. Banat, Bucureşti, Editura Academiei.
inflexional endings: that is to say that the phonological constraint, if real, is inherently circumscribed to a morphological envrionment and in that respect already inherently morphological.
Byck, J., A. Graur, 1967, “Influenţa pluralului asupra singularului substantivelor şi adjectivelor în limba română”, in: Jacques Byck. Studii şi articole, Bucureşti, Editura Ştiinţifică, 49–92.
Calotă, I., 1986, Contribuţii la fonetica şi dialectologia limbii române, Craiova, Scrisul Românesc.
Capidan, T., 1932, Aromânii. Dialectul aromân: studiu lingvistic, Bucureşti, Academia română.
Cazacu, B., 1966, “Denumirile româneşti ale ficatului şi ale plămânilor după ALR”, in Studii de dialectologie română, Bucureşti, Editura Ştiinţifică, 121–134.
Donovetsky, O., 2011, “Forme şi valori ale verbului în graiurile munteneşti (III)”, Fonetică şi dialectologie, 30, 135–179.
Gheţie, I., A. Mareş, A., 1974, Graiurile dacoromâne în secolul al XVI-lea, Bucureşti, Editura Academiei.
Iancu, V., 1965, “Despre 'natura' iotacizării verbelor româneşti”, Cercetări de lingvistică, 10, 281–288.
Iancu, V., 1966, “Vechimea iotacizării verbelor româneşti”, Cercetări de lingvistică, 11, 271–274.
Ivănescu, G., 1948=2012, “Apariţia verbelor cu -t, -d, -n şi -r”, in: Problemele capitale ale vechii române literare, Iaşi, Editura Universităţii “Alexandru Ion Cuza”, 243–267.
Lass, R., 1997, Historical Linguistics and Language Change, Cambridge, Cambridge University Press.
Lombard, A., 1954, 1955, Le Verbe roumain. Étude morphologique, I, II, Lund, Gleerup.
Maiden, M., 1996, “On the Romance inflectional endings -i and -e”, Romance Philology, 50,147–182.
Maiden, M., 2011, “Allomorphy, autonomous morphology and phonological conditioning in the history of the Daco-Romance present and subjunctive”, Transactions of the Philological Society, 109, 59–91.
Marin, M., I. Mărgărit, V. Neagoe, 1998, “Graiurile româneşti din Ucraina şi Republica Moldova”, Fonetică şi dialectologie, 17, 69–155.
Morariu, L., 1927, “Morfologia verbului predicativ român. Partea a II-a: flexiunea”, Codrul Cosminului, 2/3, 285–346.
Morariu, L., 1929, “Morfologia verbului predicativ român. Partea a II-a: flexiunea. Prezentul conjunctiv şi imperativul”, Codrul Cosminului 4/5, 1–96.
Petrovici, E., 1956-, Atlasul lingvistic romîn, n.s., Bucureşti, Editura Acadmiei.
Pop, S., 1952, “La iotacisation dans les verbes roumains”, in: Mélanges de linguistique et de littérature romanes offerts à Mario Roques III, Paris, Bade, 195–235.
Popovici, I., 1914, Dialectele române din Istria, I: Referinţele sociale şi gramatica, Halle.
Puşcariu, S., 1937a, “Essai de reconstruction du roumain primitif”, in: Etudes de linguistique roumaine, Cluj-Bucureşti, Imprimeria naţională, 64–120.
Puşcariu, S., 1937b, “Considérations sur le système phonétique et phonologique de la langue roumaine”, in: Etudes de linguistique roumaine, Cluj-Bucureşti, Imprimeria naţională, 203–259.
Puşcariu, S., 1994, Limba română. II. Rostirea, Bucureşti, Editura Academiei.
Sala, M., 1976, Contributions à la phonétique historique du roumain, Paris, Klincksieck.
Saramandu, N., 1992, “Iotacizarea verbelor şi structura dialectală a dacoromânei”, Limba română, 41, 83–87.
Saramandu, N., 2005-, Atlasul lingvistic român pe regiuni. Sinteză, Bucureşti, Editura Academiei Române.
Stan, I., D. Uriţescu, 2000-, Noul atlas lingvistic român, Crişana, Bucureşti, Editura Academiei Române.
Vasiliu, E., 1968, Fonologia istorică a dialectelor dacoromâne, Bucureşti, Editura Academiei.
Zamfir, D.-M., 2005, Morfologia verbului în dacoromâna veche (secolele al XVI-lea−al XVII-lea), I, Bucureşti, Editura Academiei Române.
