Romanian DOM and Loss of Analyzability

Abstract

This paper revisits the diachronic changes to Romanian DOM by focusing on the emergence of the DOM particle pe: the prenominal preposition pe is shown to undergo loss of analyzability when (i) the adjacent noun phrase is the direct object of the verb; and (ii) pe-DP falls under a certain pragmatic treatment. In other contexts, pe continues as a preposition. Loss of analyzability entails modification of the feature bundle associated with pe, as well as chunking and sensitivity of pe-noun phrases to discourse related priming factors. Briefly, the chunk consisting of two segments (i.e., prepositional phrase and nominal phrase: PP > DP) is gradually reduced to one segment (i.e., DP). This transition is context dependent; that is, it intensifies when the DPs receive a reading that involves discourse salience and animacy. The loss of analyzability regarding the properties of pe and the structural consequences it implied provide the basis for assessing the advent of animacy and definiteness/specificity as priming factors for DOM in Modern Romanian.

1. Introduction

Differential marking is a cross-linguistic mechanism to mark constituents for various purposes ((de)emphasis, grammaticality, thematic enhancing etc.) by using various means, such as dedicated particles, Case alternations, clitic doubling and so on (Aissen, 2003, a.o.). The option for and the pairing of purpose and means vary from language to language and also at the level of inter- and intra-language. In this paper, the discussion of differential marking focuses on the direct objects in Old and Modern Romanian.

Differential marking of direct objects (DOM) is an outstanding property of Romanian grammar that has maintained its robustness since Old Romanian (i.e., 16th–18th centuries) up to Modern Romanian (19th–21st centuries). However, the extent to which DOM occurs, its context of use and its morpho-syntactic properties have changed considerably over time. The main clue for a DOM treatment of the direct object comes from the presence of pe in prenominal position, which remains constant across stages of Romanian (free phonetic variation between pre and pe in Old Romanian, and only pe in Modern Romanian).

In Modern Romanian, DOM is realized with pe and shows sensitivity to animacy and definiteness/specificity. Moreover, its application depends on the type of nominal category: it is obligatory with tonic pronouns and proper names, as in (1a), but optional with some animate common nouns, as in (1b), and excluded with inanimate nouns, as in (1c).

(1)	a.	*(L)-am		chemat	*(pe)	el/	*(pe)	Dan.
		cl.acc.m.3sg-have1		called	dom	him/	dom	Dan
		‘I called him/Dan.’
	b.	Am	chemat	studentul.//	L-am			chemat	pe		student.
		have.1	called	student.the	cl.acc.m.3sg-have.1			called	dom		student
		‘I called the student.’
	c.	Am	chemat	taxiul.//	*L-am			chemat	pe		taxi.
		have.1	called	taxi.the	cl.acc.n.3sg-have.1			called	dom		taxi
		Intended. ‘I called the taxi.’

The Old Romanian picture is however more complex. Notably, in 16th century texts, three DOM strategies can be found: (i) clitic doubling (CD) of the noun, as in (2a); (ii) the insertion of pre in front of the noun, as in (2b); (iii) a combination of clitic doubling and pre, as in (2c).

(2)

a.

şi

nouă-lk

dezlegă

[visul

nostru]k (PO 264)

and

to.us-cl.acc.m.3sg

explained

dream.def.n.sg

our.n.sg

‘and he explained to us our dream’

b.

încă

cu

o

bătaie voiu

lovi

[pre Faraon]

şi

[pre Eghipet] (PO 480)

also

with

an

attack will.1sg

hit

pre-Pharaoh

and

pre-Egypt

’I will hit the Pharaoh and Egypt with another attack’

c.

ducă-lk

[pre acesta]k

domnu-său

înaintea

domnedzeilor

take.imp.2sg-cl.acc.m.3sg

pre-this.one

master-his

before

gods.the.gen

’let his master take this one before the gods’

(PO 545)

The CD strategy in (2a) was already unproductive in the 16th century and practically disappeared in the 17th century (Pană Dindelegan, 2016). On the other hand, the strategy in (2b) became very productive, with extension to the combination in (2c) (Hill & Mardale, 2021). In the 21st century, only the DOM option in (2c) is productive (Tigău, 2020). Since prenominal pre appears in both (2b) and (2c), our paper will focus on this element, given that the diachronic changes it underwent provide information on the emergence and evolution of DOM in Romanian.

The challenge for assessing data as in (2b, c) comes from the observation put forth in this paper that pre may not have had a DOM effect at all times. More precisely, as shown in (3), pre may or may not occur with either animate or inanimate nouns in direct object position (see also van Eeden, 1985, p. 381).

(3)

a.

de

nu

veţi

ierta

[pre

vrăjmaşii

voştri]

(ITZ 2r/13)

if

not

will.2pl

forgive

pre

enemies.the

your

‘if you will not forgive your enemies’

b.

Iubiți

[vrăjmaşii

voştri].

(ITZ 3r/18)

love.imp.2pl

enemies.the

your

‘love your enemies’

c.

Cinstiţi

[pre lemnul]

de la

care

lemn

se face

ajutoriul.

(ITZ 37v/5)

honor.imp.2pl

pre cross.the

from

which

cross

refl makes

help.the

‘honor the cross from which aid comes out’

d.

Veniţi

să

băgăm

[lemnul]

în pâinea

Lui

(ITZ 30v/5)

come.imp.2pl

sbjv

insert

cross.the

in bread.the

his

‘come to insert the cross in his bread’

e.

Şi

Hanaan

rodi

[pre Sidon],

născutul

lui

dentâiu

(PO 39)

and

Hanaan

begot

pre Sidon

born.the

his

first

‘and Hanaan begot Sidon, his first born’

f.

Şi

Arcfasat

rodi

[Salah];

şi Salah

rodi

[Eber]

(PO 39)

and

Arcfasat

begot

Salah

and Salah

begot

Eber

‘and Arcfasat begot Salah, and Salah begot Eber’

In the Romanian of the 21st century, the alternation between +/−DOM, as in (3), is restricted to some quantifiers and common nouns with human reference, as DOM became obligatory with all pronouns and human names. That is, part of what was an optional process in the 16th century became obligatory in the 21st century Romanian grammar.

Notably, the theories of DOM identify animacy and/or definiteness/specificity as the underlying factors common to all DOM strategies, cross-linguistically (de Hoop & Malchukov, 2007; Malchukov & de Swart, 2008, a.o.). From this point of view, the examples in (3) are problematic: pre with inanimates, as in (3c), may signal that this is not a DOM construction, while lack of pre with definite/specific nouns, as in (3b, f), signals that definiteness/specificity does not count for DOM in Old Romanian. In fact, the diachronic impact of animacy and definiteness/specificity on the incidence of DOM in Romanian has not received much attention so far. This paper will show that animacy gains importance for the emergence of DOM readings during the process by which the prepositional analysis of pre becomes gradually superfluous when the noun is in direct object position; that is, direct objects with animate nouns are the first to display DOM as the pe-noun phrase is analyzed as a DP instead of a PP.

Definiteness/specificity becomes a relevant factor for DOM at a later diachronic stage, according to the data presented in this paper. For example, in 17th century texts, names of persons, inherently animate, display pre for DOM almost systematically, whereas animate nouns with specific reading do so sporadically, e.g., ‘saint’ without pre in (4a) but with pre in (4b).

(4)

a.

cu

glas

groznic,

dzise

să

bată

[svântul]

preste gură (Var 74v)

with

voice

loud

said

sbjv

beat.3

saint.the

over    mouth

‘he said with a loud voice to slap the saint over his mouth’

b.

dzise

slujitorilor

să

ia    [pre svântul] in suliţe (Var 73r)

told

servants.the.dat

sbjv

take.3  pre

saint.the in  spears

‘he told the servants to lift the saint in spears’

The diachronic trajectory established in this paper when it comes to animacy, in relation to pre-direct objects, is as follows: (in)animate (16th c.) → animates (17th/18th c.) → humans (19th c.). That is, direct objects containing animate nouns show a gradually increasing tendency to be introduced by pre, indicating that animacy becomes a trigger for DOM. However, for the same noun class, definiteness/specificity does not qualify as a systematic trigger for DOM in Old Romanian, although it does so in Modern Romanian, in certain contexts (e.g., with complex noun phrases).

The new perspective proposed in this paper on data as in (2) to (4) throws doubt on the systematic use of pre as a DOM particle in the 16th century texts, or in the archaic language register of some religious texts of the following century. Briefly, the proposal is that as long as pre qualified as a functional preposition, it only served as an accusative case marker for direct objects. The DOM property of pre emerges when it becomes an element of the nominal phrase (DP) instead of projecting a separate prepositional phrase (PP), that is, when the case assigning property of the preposition is lost (as pre ceases to function as a lexical preposition).1 Animacy as a priming factor for pre marking arises concurrently with this morpho-syntactic change.

Technically, the analysis proposed in this paper relates the diachronic changes of DOM to the loss of analyzability arising from changes in the feature composition of the syntactic elements involved in DOM, especially regarding the properties of pre. The instability of the featural composition resulted in synchronic variations in the treatment of DOM in Old Romanian, and this variability was channeled through a gradient process towards a conventionalized, more stable system for DOM in Modern Romanian.

2. Background

2.1. Theoretical Framework

The emergence of DOM in Romanian will be assessed in terms of animacy and definiteness/specificity scales. It is well established, at least empirically, that languages tend to have privileged categories which must be differentially marked, such as personal pronouns or proper names. The status and organization of such categories indicate a correlation with features such as specificity, definiteness, animacy or topicality. Extensive research since Silverstein (1976) or Aissen (2003) among others, has identified typologically common hierarchies such as the ones in

Table 1. Scales for differential object marking.

a. Animacy/person: 1/2 pronoun > 3 pronoun > proper name > human > animate > inanimate

b. Specificity/definiteness: pronoun > name > definite > specific indefinite > non-specific

c. Topic accessibility scale: active > accessible > unused > brand-new anchored > brand– new unanchored

below, taken from Dalrymple and Nikolaeva (2011). Generally, the higher an object is on the scale, the higher the differential marking probability (Silverstein, 1976; Lambrecht, 1994; Lazard, 2001; Aissen, 2003; Næss, 2004; Dalrymple & Nikolaeva, 2011, a.o.). The scales predict, for example, that a language displaying DOM with names of persons necessarily displays DOM with strong personal pronouns as well. In addition, discourse factors have also been detected as triggers for DOM cross-linguistically, especially where various types of topic interpretation are concerned (Dalrymple & Nikolaeva, 2011; Iemmolo, 2010; Leoneti, 2008, a.o.).

DOM in Modern Romanian displays sensitivity to animacy and definiteness/specificity, but this was not always the case. The paper discusses the stages through which Romanian ended up complying with the scales in Table 1, and the role topicality played in this process.

The analysis proposed in this paper relies on the conceptual contrast between reanalysis and loss of analyzability, such as developed and applied in Bybee (2003, 2010); Torres Cacoullos (2015); Delicado Cantero and Amaral (2024) among others: while reanalysis involves a sudden change in the categorial status of an item during the process of language acquisition (a learner acquires grammar G2 instead of G1; e.g., Roberts & Roussou, 2003), loss of analyzability involves a gradual process arising from the learner’s usage of the language, with repetitions and changes in the chunking of the lexical material. A typical example comes from the change from going to to gonna in English: reanalysis would entail that a learner reanalyzed [goingto] upward in the clause hierarchy from being merged as the verb go to being merged as a compound modal or auxiliary. On the other hand, loss of analyzability entails that the composite morphemes in going to lose their association with go, to or -ing (Bybee, 2003) in certain contexts of use, while being preserved in other contexts.

Loss of analyzability is a gradual process that involves priming factors. A crucial concept in psycholinguistics (psychology more generally), priming is related to how exposure to one stimulus may unconsciously influence or modulate the response to a subsequent (related) stimulus (Weingarten et al., 2016). As Bock and Griffin (2000, p. 17) put it in a classical definition, “the unintentional and pragmatically unmotivated tendency to repeat the general syntactic pattern of an utterance” is at the core of priming. Important streams of work in diachronic linguistics have emphasized the non-trivial role of persistence in the creation, development, and expansion of new structures.

From this perspective, it appears that the changes to the DOM system in Romanian, such as attested between the 16th to the 21st centuries, have been, so far, discussed only in terms of reanalysis (Mardale, 2015; Hill & Mardale, 2021; Tigău, 2011, a. o.). Approaching the same material by considering the diachronic changes of Romanian DOM as loss of analyzability may strengthen the explanatory power of the attested variation at least with regard to two aspects: (i) It predicts that the same learner may concurrently use pe as a preposition and as a DOM particle (i.e., instead of a complete parametric switch from a grammar with pe as a preposition to a grammar with pe only as a DOM particle); (ii) It allows for the gradual and independent introduction of animacy and definiteness as criteria for DOM in this language under context dependent conditions, with non-trivial variations from one text to another. The latter changes occur under priming factors related to discourse manipulations.

2.2. Corpus and Methodology

The discussion developed in the present paper relies on the data gathered from 16th century religious translated texts and 17th century texts which were either translated or written directly in Romanian. The texts we have consulted are discussed in detail in Section 5 and the full list is provided in the References section. Out of this corpus, six texts were searched for tokens of DOM: two translations from the 16th century (PO and EV), and four texts from the 17th century. The latter include two short Chronicles (CM and Ur) and two religious texts for comparison (BB and Var). In addition, we consulted a text from the 19th century (IC) and official grammars of Modern Romanian for the 20th and the 21st centuries.

Philologists debate the reliability of the 16th century translated texts for assessing the development of DOM in Romanian. For example, Densusianu (1900/1997) considers that the translators sacrificed the Romanian grammar in order to respect the syntax of the original. The reason why DOM was rare in 16th century translations is that the language of the original did not have it. On the other hand, Puşcariu (1921–1922) argues that the use of pre-DP direct objects was emergent and unstable in the 15th–16th centuries, which explains why it appears sporadically and even with inanimate nouns (which were unmarked in the language of the original text). Current historical linguistic studies implicitly adopt Puşcariu’s path, while exercising caution in the selection of texts (Pană Dindelegan, 2016; Stan, 2013). For this paper, the text selection follows the recommendations from such studies.

The methodology adopted in this paper relies mainly on considering the contexts in which the same noun phrase does or does not fall under DOM, as a way of measuring the rate of increase in DOM application. While statistic measurements of such variations are already available in the literature (e.g., von Heusinger & Onea, 2008; Avram & Zafiu, 2017; Tigău, 2020; Hill & Mardale, 2021), the contextual conditioning for such variation has been less discussed.

In this paper, the loss of analyzability will be determined by means of syntactic tests establishing categorial unithood. For example, given a certain syntactic context (e.g., direct object position), what is the transition from the analysis of pre-nouns as consisting of two phrases (i.e., PP > DP) to their analysis as one phrase (i.e., DP). In other words, we establish the gradual change in the chunking and recategorization of the DOM-ed constituent. The priming condition for the analysis of pre as a nominal marker for DOM instead of a preposition is shown to be animacy and discourse manipulation; e.g., whether the speaker intends to make the direct object salient or not.

Technically, the loss of analyzability will be determined in this paper by adopting a feature-based approach to syntactic derivation (Chomsky, 2008, et seq.). More precisely, each syntactic node maps certain formal features which determine the syntactic status of a lexical item. For example, relevant to DOM is the type of features realized under the node in which pre merges. When merged as P, pre may have substantive features or only functional features, and in both instances, it also has a [case] feature (i.e., accusative) to impart to the noun phrase. However, when merged as D (i.e., new chunking), pre loses the [case] feature, and the noun phrase has to receive the accusative case differently (i.e., from the functional structure of the verb, not from pre). Along these lines, loss of analyzability is reinterpreted as a gradual and context-dependent loss, modification or variation with respect to the featural composition of the syntactic nodes.

3. PE-DP Internal Structure

The particle pre is fundamental for the implementation of DOM in Romanian. Hence, this section considers the concurrent use of pre as a fully fledged preposition and as a functional preposition for nouns and pronouns (henceforth, DPs) in direct object position. Briefly, the proposal is that up to the 17th century, pre projected a prepositional phrase (PP) structure in both syntactic positions although its feature composition was reduced when it introduced complements. Crucially, the data of religious 16th century texts are inconclusive when it comes to establishing whether pre preceding direct objects had the function of a case marker or of a DOM particle in the religious translations dated for that century.

3.1. From Lexical to Functional P

As a substantive preposition, pre has location, temporal, aboutness semantics, and selects DPs to which it assigns accusative case, as in (5). These properties of pre are well preserved up to the 21st century, although slight changes in the semantic features have occurred (Mardale, 2015); compare Old Romanian (5a) with Modern Romanian (5b).

(5)	a.	toţi pomii	în câmp,	carii	mainte	nu	era	pre pământ		(PO 15)
		all trees.the	in field	which.the	before	not	were	on earth
		‘all the trees of the field that previously have not existed on earth’
	b.	toţi pomii	din câmp,	care	mai înainte		nu	erau	pe pământ
		all trees.the	in field	which	more before		not	were	on earth
		‘all the trees of the field that previously have not existed on earth’

In constructions as in (5), the PP-pre is an adjunct (i.e., the theta-role it lexicalizes is not obligatory in the theta-grid of the verb) versus selected (i.e., obligatory). The features associated with pre are a substantive semantic feature [SEM] (e.g., location), and a formal [case] feature, which support the syntactic configuration in (6).

(6)

At unattested times and for debatable reasons, PP-pre emerged as a direct object. Philological studies (see Sala, 2001 for an overview) relate the use of pre with direct objects to the gradual loss of inflectional nominative and accusative case marking in the transition from Latin to Romanian. More precisely, since changes of word order from SVO to VSO happened concurrently with the loss of inflectional endings in nouns, prenominal pre provided indications w.r.t. which post-verbal noun is the subject or the object in a sequence as in (7): who is the agent, the ‘man’ or the ‘lion’? This rationale led to hypothesis (7’).

(7)	Prinde	omul	un	leu
	catches	man.the	a	lion
	‘The man catches a lion.’ Or
	‘A lion catches the man.’
(7’) Hypothesis: pre is an accusative case assigner to nominal direct objects.

Hypothesis (7’) has been disputed on the basis of solid evidence; for example, strong pronouns have preserved the accusative case inflection up to the present (e.g., eu ‘I.nom’ vs. mine ‘I.acc’) and yet they display pre in direct object position (Mardale, 2015, a.o.). Instead, other explanations have been entertained, such as a reanalysis of the PP-pre with aboutness meaning from an adjunct to a selected position, in contexts with certain verbs (Onea & Mardale, 2020); see (8). This proposal led to hypothesis (8’).

(8)

a.

se-au

tocmit

[pre

megiiaşi]

(DRH, 1671)

refl.3-have.3

haggled

about

neighbours

‘They haggled over their neighbours’

b.

amu

tocmit

[pre Muşat

postelnic]

şi

have.1

hired

pre Mușat

chamberlain

and

[pre Negoe]

pântru

4

rumâni

vii

şi

pântru

pre Negoe

for

4

serfs

alive.m.pl

and

for

6 delniţi,

fârî

rumân (DÎ, 1622)

6 plots.of.land

without

serf

‘I hired Musat the chamberlain and Negoe in exchange for 4 living serfs and for 6 plots of land, without serfs’

(8’) Hypothesis: pre introduces a pragmatic role in the theta-grid of the verb.

Onea and Mardale (2020) point out that pre-DP in (8a) conveys the topic of the negotiation in relation to the reflexive verb tocmi ‘haggle’. In (8b), the pre-DP is reanalyzed as the direct object of the same verb, coinciding with a change in meaning from haggling to hiring, which also involves a degree of negotiation.

The main difference between the two hypotheses consists in the type of features associated with pre: For the hypothesis in (7’), pre changes from a lexical to a functional preposition, i.e., it loses [SEM] and preserves [case]. For the hypothesis in (8’), pre-DP changes the merging position on the clausal spine without inferring any modification to the internal structure (i.e., pre is still [SEM], [case], having an aboutness feature). So, the justification for positional change lies with the verb, which modified its theta-grid: it became transitive in (8b) (it is reflexive in (8a)) by adding an obligatory aboutness theta-role (an E-topic in Onea & Mardale, 2020). Thus, DOM is triggered as the verb’s E-topic gets checked against the [SEM] of pre. Considering that topicality is a contributor to the application of DOM, as signaled in Table 1, the reanalysis of PP pre-DP as a DOM construction conforms to the cross-linguistic patterns.

Nevertheless, from the perspective of the economy principle, hypothesis (7’) is more economical than hypothesis (8’), and also easier to support with the data. Although the proposed explanation for using pre as a differentiator between subjects and objects as in (7’) can be indeed dismissed, it is however obvious that pre was used as a replacement for accusative endings at some stage. Ledgeway (2012) points out a generalized tendency of Romance languages to use ad or per as case markers for accusative during the transition from Latin to Romance, irrespective of the SVO or VSO orders. Romanian seems to be no exception, and considering that this is a conservative language, with a slow pace of change, it would not be surprising to find left-overs of that process even in 16th century texts. The next section presents evidence in this respect.

3.2. The Case Marker Pre

An explicit definition of pre as a case marker for direct objects is found in Eustatievici’s grammar (1757). Eustatievici Braşoveanul was a school teacher who had received a religious education (his father was a priest) and had worked with religious translations using versions that must have predated the copied versions preserved up to the present. In other words, he was familiar with a more archaic grammar of Romanian than the one reflected in the spoken language of the 18th century. He was teaching these texts in religious schools, so the prescriptive grammar he wrote to serve him as a textbook must have catered to the archaic register of the religious texts, where direct objects with pre-DP did not conform to DOM.

Briefly, Eustatievici explicitly defines pre in these contexts as a case marker for accusative, in the same way the ending -lui inflects the noun for genitive (thus, creating the form lucrului below). His case marking paradigm is reproduced in (9).

(9)

N. acest lucru;

lucrul   acesta.

N.a

acestuia

lucru;

-ului

-sta.

this thing

thing.the this

gen

this.the.gen

thing

-the.gen

-this

D. acestui

lucru;

-ului

-sta.

P. pre

acest

lucru;

-ul –sta.

this.dat

thing

-the.dat

this

acc

this

thing

the this

C. o,

lucru,

o,

lucrule

L. de la acest

lucru;

-ul

-sta. (E {34},<42/30v>)

oh

thing.voc

oh

thing.the.voc

of at this

things

the

this

‘Nom. this thing; Gen. of this thing; Dat. to this thing; Acc. this thing; Voc. oh thing! Loc. from this thing’

Accordngly, the language of his grammar uses pre systematically with direct objects irrespective of their semantic classification, as in (10) (see also Hill & Mardale, 2021).

(10)

a.

cel

ce

după

sine

dobândeşte

pre

căderea

pricinuitoare

that

which

after

itself

obtains

acc

fall.the

causal

‘the one that, after itself, obtains the Case inflection’ (E {48}<66/42v>)

b.

schimbând

pre

vremea

cea

trecută

(E {69}<98/58v>)

changing

acc

time.the

the

past

‘changing the past tense’

c.

primind

pre

părticelile

sale

(E {69}<99/59r>)

receiving

acc

particles.the

its

‘receiving its particles’

d.

carele

rânduieşte

pre

fieştecare

parte

a

cuvântului

which.the

organizes

acc

each

part

gen

word.the.gen

‘which organizes each part of the word’ (E {90}<134/76v>)

Vestiges of this function of pre appear in religious texts such as VCR-Voronet Codex (1563–1583), ITZ (1642) and BB (1648), where the direct objects it introduces may or may not comply with DOM (see also Stan, 2013). More precisely, the noun in pre-DP could be +/− definite and/or +/− animate, in a way that matches Eustatievici’s rule. Furthermore, the syntactic behavior of these pre-DP constructs indicates a PP structure that is typical of case assigning prepositions. The difference from Eustatievici’s prescriptive grammar is that, in the above mentioned texts, direct object DPs also follow the verb directly, without a pre connector. Illustrations from these texts are provided and discussed below.

3.3. Tests for the PP Structure

The suggestion arising from the data in (9) and (10) is that pre maintains a prepositional status, but its semantic input is lost, whereas the case properties are preserved. In other words, the internal structure was still as in (6), only the [SEM] feature of P is lost. Evidence for this structure comes from two types of data:

(i)

PP-pre as direct object to nominals

It is an established generalization that nouns and adjectives cannot assign case to the nouns in their DP objects, so the noun either comes inflected for case or is preceded by a functional preposition that assigns it case (Kayne, 1994, a.o.); hence, the contrast seen in Modern Romanian between (11a) with genitive case or (11b) with the preposition de versus the ungrammatical (11c) with none of the above.

(11)	a.	iubitoarea	[animalelor]
		lover.def.f.sg	animals.def.m.pl.gen
		‘the lover of animals’
	b.	iubitoarea	[de animale]
		lover.def.f.sg	of animals
		‘the lover of animals’
	c.	*iubitoarea	animale
		lover.def.f.sg	animals

In the 17th century texts, all the options in (11) can be found. In addition, the options in (12a, 12b, 12c) also occur, where pre has the same function as de in (11b).

(12)

a.

aceasta

pasăre

iaste

iubitoare

foarte

[pre

soţulŭ

său.]

(van Eeden, 1985, p. 383;

this

bird

is

loving

very

pre

spouse.the

her

VCR 421/29-30)

‘this bird is very enamored of her spouse’

b.

de

iaste

neştine

cinstitoriu

[pre Dumnedzău]

(van Eeden, 1985, p. 383; VCR 121/25-6)

if

is

someone

respectful

pre God

‘if someone is respectful of God’

c.

Dumnezeu

iaste

cunoscătoriu

de inimi.

(van Eeden, 1985; ITZ 22y/4-5)

God

is

knowledgeable

of hearts

‘God is an expert of hearts’

The alternation between de and pre as functional prepositions in these contexts is excluded in Modern Romanian, where only de is grammatical. This means that the analysis of de has not changed in these contexts, whereas the analysis of pre did, in a way that makes it unable nowadays to assign the accusative case to selected nouns.

(ii)

Modifiers to PP-pre

In principle, the structure in (6) must allow for phrasal elements to merge in Spec, PP, which can be seen with cam ‘almost’ in the PP adjunct in (13), in Modern Romanian.

(13)	Stă [cam	pe	marginea	patului.]
	sits almost	pe	edge. def.f.sg	bed.def.n.sg.gen
	‘s/he sits almost on the edge of the bed’

This possibility is attested in Old Romanian (versus Modern Romanian) with pre-DP in object position, as in (14), where singur is interpreted as an intensifier of the pronoun introduced by pre.

(14)	să	iubeşti	pe vecinul	tău	ca	[singur	pre tine].  (ITZ 54v/1)
	sbjv	love.2sg	pe neighbour. def.m.sg	your  as		self	pre you
	‘love your neighbour as you yourself’

The data in (12–14) provide evidence that the merge of pre-DP in a complement instead of an adjunct position involves loss of analyzability insofar as no semantic features are associated with pre introducing a direct object DP. However, the [case] feature is preserved, which is sufficient to allow for pre to project a PP with the structure in (6). There is no evidence that in such contexts pre would have DOM effects.

4. Chunking for DOM

The discussion of the data so far pointed out that, in Old Romanian, pre could introduce DPs either as adjuncts or as complements, but its features differed according to the syntactic position: loss of analyzability happens only in the complement position, which converts the lexical preposition to a functional one. This change was paired with the loss of declension for accusative case during the romanization process, in the transition from inflectional to analytical case systems.

However, the analytical case system in Romanian is relevant only to adjuncts and partially to indirect objects, insofar as the preposition became obligatory for case assignment to DPs in such syntactic contexts; so PP-pre has been well preserved as an adjunct. When it comes to direct objects, verbs assign structural case to them, dispensing with functional prepositions. Accordingly, loss of analyzability for pre (i.e., loss of [case]) with direct object DPs must be expected once structural case is stabilized: the functionally redundant pre is then either dropped or recycled for other use, since it is dissociated from both [SEM] and [case].

The 16th century religious/translated texts reveal significant inter- and intra- speaker variation in the use of pre-DPs in direct object position. Importantly, at this diachronic stage, pre-DP alternates with the occurrence of the same DP by itself as a direct object, which signals the stabilization of structural case assignment by verbs. However, the alternation in the use of the direct object with or without pre seems to be aleatory, as no systematic priming factors for the use of pre in this syntactic context can be identified. This contrasts with the situation of the 17th century texts, where the use of pre as a differential object marker is obvious and more predictable according to the discourse background and animacy.

In light of these observations, it is difficult to assess whether in the 16th century texts or in the religious texts with archaic register (e.g., BB) pre was still used as a case marker (i.e., as a functional preposition), as found in the examples of the previous section, or only as a DOM particle. Possibly, both uses concurred to some extent, which explains the unsystematic use of this element with direct objects. Consider the contrasts in the minimal pairs in (15).

(15)	a.	Văzu-o		[ea] Isus, strigă	şi	zise	ei	(BB 454)
		saw.3sg-cl.acc.f.sg		her Jesus called.3sg	and	said.3sg	her.dat
		‘Jesus saw her, called and said to her’

	b.	Văzu	[pre ea] Isus,	strigă	şi	zise	ei	(BB 458)
		saw. 3sg	pre. her Jesus	called.3sg	and	said.3sg	her.dat
		‘Jesus saw her, called and said to her’

	c.	Şi văzură	[pre	el]	ucenicii	îmblând	pre	mare	(BB 293)
		and saw.3pl	pre	he	apprentices	walking	on	sea
		‘and the apprentices saw him walking on the sea’

	d.	Şi	văzură	[el]	ucenicii	pre	mare	îmblând	(BB 290)
		and	saw.3pl	he	apprentices	on	sea	walking
		‘and the apprentices saw him walking on the sea’

	e.	că leagea	amu	den	tocmeală	slugă	are		[pre frica]	(EV 16)
		that law.the	now	from	judgment	servant		has	pre fear.the
		‘that now the law the fear has as a servant in the judgment’

	g.	potoli	Hristos [spământarea		şi	frica]	ce	avea	apostolii, (EV 137)
		calmed.3sg	Christ	scare.the	and	fear.the	that	had	apostles.the
		‘Christ calmed the scare and the fear that the apostles had’

In (15), the use of +/−pre occurs in identical pragmatic and syntactic contexts, for animate and inanimate nouns. The strong tendency is, however, to find pre with strong pronouns and names of persons rather than with other nominal classes. For example, in EV, the verb forms avea ‘have’ and avut ‘had’ appear with 32 direct objects: two out of five animate nouns are preceded by pre, compared to two out of 25 inanimate nouns. So, one may conclude that animacy began to count for the option for pre, signaling the emergence of DOM.

Along these lines, the emergence of DOM through the pre marking of direct objects, as attested from 17th to the 21st centuries, has to be understood as resulting from the change in the formal features associated with pre under two conditions:

(i)

The stabilization of structural case assignment. This led to the loss of analyzability of pre as a preposition in selected contexts, i.e., the general loss of [SEM],[case].

(ii)

The combination of animacy and discourse manipulation (see the discussion in Section 5). Pre is dissociated from the [case] feature and associated with a discourse feature [d] that conferred salience to the direct object (e.g., through topic or focus treatment). Animate nouns were more likely to qualify for a salient reading due to their high degree of referentiality.2

Loss of analyzability along these lines had an impact on the internal structure of the pre-DP construct. That is, pre ceases to project a PP as in (6), and is merged, instead, in the DP field. In other words, the PP > DP structure is chunked to DP only, although pre is still present. As shown in (16), pre merges at the high left periphery of the DP where discourse features have been shown to be mapped, independently of DOM (Giusti, 2005, a.o.).

(16)

According to the discussion so far, [d] in (16) is a formal feature that maps both animacy and salience, and whose exact interpretation arises from a combination of the noun semantics, the syntactic position and the pragmatic context. The structural change from (6) to (16) means that DP₁ in (16) must receive structural accusative case from the verb (not from a preposition).

The tests for chunking and reduction of a PP as in (6) to a DP as in (16) rely on Modern Romanian data, to show the most recent results of the diachronic change. More precisely, the tests in (17) to (19) indicate that pre ceases to function, at some point, as a case assigning preposition for direct objects, although it is preserved as prenominal.

(i)

Complements to nouns and adjectives

Examples as in (12) became ungrammatical in Modern Romanian: pe can no longer introduce direct objects to nouns and adjectives. Only de serves this purpose, as shown in (17).

(17)	iubitoare		de copii	versus	*iubitoare	pe copii	versus	iubeşte	pe copii	(Modern Romanian)
	lover.f.sg.	of children			lover.f.sg	pe children		loves	pe children

In (17), pe can precede the DP object of a verb, but not of a noun. This is possible because the direct object does not depend on pe to assign it case under selection by a verb (i.e., the verb assigns the structural case to the entire pe-DP) but it would do so under selection by a noun. Hence, pe ceased to be analyzed as a functional preposition, which indicates the loss of the [case] feature.

(ii)

Optionality of pe with CLLD

Prepositions, be they lexical or functional, obligatorily precede the DP when constituent movements to preverbal positions occcur. This is expected as long as the DP depends on its case assigner (here, P) to be recognized as a syntactic item. Notably, in Modern Romanian, pe is preferred but not obligatory upon the CLLD of the direct object, as in (18).

(18)

a.

[*(De) animale]

e

mare

iubitoare.

of  animals

is

great

lover

‘She is a great lover of animals.’

b.

[*Patul/*(Pe)

pat] nu

e nimic.

bed.the/on

bed neg

is nothing

‘There is nothing on the bed.’

c.

[Copiii/  Pe copii]

nu

i-am

invitat.3

children.the/dom children]

neg

cl.acc.3pl-have.1

invited

‘I did not invite the children.’

d.

Toţi,

nu

i-am

mai

chemat

la

mine.

all.m.pl

neg

cl.acc.3pl-have.1

more

called

at

me

‘All, I haven’t invited them to my place after all.’

The contrast between (18a, 18b) versus (18c/d) is that in (18c, 18d) pe is not the source of case for the DP under CLLD, hence its optionality. The DP copy left under the verb ensures the DP’s compliance with the case requirement. On the other hand, CLLD entails pre-DP almost systematically in 16th century texts, as detailed in Section 6. The inference is that the status of pre as either P or D was ambiguous in the 16th century, but not so in the 21st century.

(iii)

Optionality of pe in answers

(19)	a.	Pe	cine		a invitat?
		dom-p	whom		has invited
		→	Băieţi		şi fete.//	Pe	băieţi	şi	pe	fete.
			boys		and girls	dom-p	boys	and	dom-p	girls
		‘Whom did s/he invite?’ → ‘Boys and girls.’
	b.	Unde	a	pus	florile?	→	*(Pe)	o masă.
		where	has	put	flowers.the		on	a table
		‘Where did s/he put the flowers?’ → ‘On a table.’

The test in (19) shows that, for Modern Romanian, when pe is a preposition introducing the noun as an adjunct, it is obligatory in the short answer, as in (19b). On the other hand, pe is optional with the direct object in the short answer in (19a). This signals that pe in (19a) does not function as a preposition, or else it would also be obligatory.

Consequently, the formal account is that the presence of pe with direct objects in the 21st century is due to a direct merge of this element at the left periphery of the DP (vs. PP), that is, in the derivational space dedicated to the mapping of discourse/conversational features to syntax, as in (16). This configurational change, from (6) to (16), follows from the increased association of pe with discourse features when its relevance as a case marker has been gradually lost. The transitional process from preposition to DOM particle (and from PP > DP to DP) involved synchronic inter- and intra-speaker variation such as attested in the texts. The relevance of the discourse feature as a priming factor for the stabilization of pe-DOM particle is discussed in the next section.

5. Priming Factors

Section 3 and Section 4 showed the starting and the ending status of pe, from preposition to DOM particle. This section identifies the factors that contributed to the use of pe as a DOM particle, as well as the triggers for DOM in its emerging stages. Briefly, the priming factors for the emergence and the gradual stabilization of DOM in Old Romanian are discourse salience and animacy: while animacy is sufficient for DOM with pronouns and persons’ names, it is not so with the other nominal categories. The latter display DOM only in certain discursive contexts that involve a salient reading.

5.1. Tokens

This section lists the tokens of pe-DP in the selected texts of the 16th and 17th centuries. The cases of DOM through CD alone, which was illustrated in (2a), will be discussed separately in Section 6. The tokens discussed in this section are relevant to three morpho-syntactic strategies shown by direct objects: unmarked, differentially marked with pre and differentially marked by both pre and accusative clitic doubling. We have extracted all the direct objects encoded through these strategies in the six texts researched. For the unmarked objects, we took into consideration just the animates, as these classes are precisely the ones which end up acquiring the differential marker in later stages. Furthermore, for all the classes of animate nouns, we only counted the post-verbal direct objects, to ensure accuracy regarding the selection of a DOM-ed DP as opposed to fronted DPs that may display pre and/or the clitic for a different reason than DOM.

The results are summarized in

Table 2. DOM in 16th and 17th centuries (animate direct objects).

Verbs	PO + Ev	BB	CM	Ur	Var
Aduce ‘bring’	31-40-10	4-37-0	5-11-0	1-6-1	7-12-1
Alege ‘choose, elect’	2-2-0	4-11-3	3-11-0	2-4-0	2-1-1
Apuca ‘snatch, grasp’	0-3-4	0-2-2	1-2-0	0-2-0	0-1-3
Asculta * ‘listen to’	2-19-3	0-26-10	0-1-0	0-0-1	4-3-0
Bate ‘beat’	6-8-1	0-20-9	0-6-0	0-13-1	0-5-0
Chema ‘call’		2-57-13	1-15-0	3-10-1	12-9-6
Da ‘give’	48-48-5	8-41-27	2-1-0	1-4-0	4-0-3
Duce ‘lead, take’	8-18-5	0-3-6	0-0-0	0-2-0	6-5-7
Goni ‘chase’	18-13-2	0-11-11	2-9-3	0-8-2	4-4-3
Îngropa ‘bury’	7-7-1	5-19-9		0-1-1	2-4-0
Iubi ‘love’	15-15-1	0-24-10	2-4-1	0-1-1	6-10-4
Lăsa ‘let’	25-23-4	5-3-1	2-6-1	0-8-0	15-10-4
Lăuda ‘praise’	0-4-1	0-3-1	0-1-0	0-1-0	0-27-0
Lovi ‘hit’	1-5	1-1-14	0-6-0	0-3-0	0-3-0
Lua ‘take’	12-3-1	14-27-20	2-8-1	2-5-0	2-9-0
Naşte * ‘give birth to’	16-17-	14-23-1		0-2-0	7-7-1
Omorî ‘kill’	3-9-4	0-7-14	1-19-3	2-11-0	1-3-2
Pierde ‘loose’, ‘kill’	6-6-0	2-19-0	2-2-0	0-1-1	2-5-2
Prinde ‘catch’	12-3-1	0-6-3	0-6-0	0-14-1	2-4-1
Pune ‘put’	24-16-1	3-2-0	2-4-0	0-12-0	4-6-0
Ridica ‘raise’	4-3-1	0-8-5	1-3-0	0-8-0	1-0-0
Scoate ‘take out, pull’	16-9-1	5-28-42	0-7-1	0-8-0	5-7-3
Trimite ‘send’	3-9-2	4-23-24	5-22-0	2-26-0	33-15-5
Vedea ‘see’	51-28-4	9-37-16	0-3-0	0-4-0	9-15-1
Ucide ‘kill’	2-9-1	0-21-11	1-4-0	1-2-0	0-6-2

* A transitive predicate in Romanian.

for a selection of transitive predicates. The selection depended on the presence of these verbs in all six texts. The figures in each column are organized as follows: unmarked direct object as the leftmost figure; pre marked direct object in the middle; pe marked and clitic doubled direct object rightmost. For example, in PO + EV the predicate aduce ‘bring’ presents 31 unmarked animate direct objects, 40 pre marked animate direct objects and 10 animate direct objects that are both pre marked and clitic doubled.

The discussion to follow focuses on three important tendencies in the encoding of direct objects diachronically: (i) the increase of pre-DP with direct objects; (ii) the pre-DP increase correlates with animacy; (iii) across the three stages illustrated in the paper, definiteness and/or specificity is/are not a priming factor for pre-DOM in Old Romanian (as opposed to Modern Romanian, where these features are indeed a priming factor).

5.2. Animacy and Salience in the 16th Century DOM

The results of Table 2 are assessed in terms of the scales presented in Table 1. In particular, Table 2 shows a clear divide between the PO + Ev stage (16th century) and the other texts (from 17th century) when it comes to the presence of unmarked animate nouns as direct objects. According to Table 2, in the 16th century texts, various classes of animate direct objects which in subsequent stages of the language need obligatory DOM, are not only possible unmarked, but this strategy is actually quite robust. Thus, strong pronouns, which dominate the animacy scale in Table 1, may appear without DOM in PO + Ev, as further shown in (20a, 20b), and so do the names of persons, as in (20c, 20d). Not only animacy, but also definiteness/specificity is ignored as a DOM factor, as seen in (20d, 20e, 20f), where the definite article (20d), possessive modification (20e) or the use of a demonstrative without the noun (20f) appear without pre. Crucially, the incidence of pe-DPs with all these categories is aleatory, as shown in (20e) where pre is optional with animate direct objects even under coordination.

(20)

a.

Slobodzi

[el]

Domnul

Domnezeu

afară den

raiul

dulceţiei

(PO 21)

chased.3sg

he

Lord

God

out   of

heaven.

the sweetness.the

‘God chased him out of the sweet heaven.’

b.

Şi mearseră

ucenicilor

lui,

rugară

[el] grăind: (Ev 359)

and went.3pl

apprentices.the.dat

his

beseached.3pl

he saying

‘and they went to his apprentices, they beseeched him saying’

c.

în zi

în care

făcu

Domnedzeu

[Adam] (PO 24)

in day

in which

made.3sg

God

Adam

‘the day in which God made Adam’

d.

cu aceastea

să

întărâtăm

noi [Dumnezeu] spre

noi şi

să

with these

sbjv

annoy.1pl

we God     toward us and

sbjv

mâniem

[nezlobivul

şi

în-lung-răbdătoriul

Domnul]

(EV 2)

enrage.1pl

innocent.the

and

long-patient.def.m.sg

God

‘with these we may get God annoyed with us and we may enrage the unblameable and much patient God’

e.

Aşa luo

Avram

[muiarea

sa]

Sarai şi [pre Lot],

ficiorul frate-său (PO 43)

thus took.3sg

Avram

wife.def.f.sg

his.f.sg

Sara and dom Lot

son.the brother-his

‘thus Avram took his wife, Sara, and Lot, his brother’s son’

f.

Dusără [acesta],

ce

era oarecând

orb,

cătră

farisei,

(Ev 185)

took.3pl this.one

who

was once

blind.m.sg

toward

Jews

‘and they took this one, who was once blind, to the Jews’

All the direct objects in (20) require DOM in Modern Romanian, which signals a complete change in the observance of DOM scales since the 16th century. In fact, at the time of PO + Ev, certain transitive predicates allow for a higher number of direct objects without rather than with pre marking. Among these predicates are: vedea ‘see’ (51 unmarked vs. 28 marked), scoate ‘take out, pull’ (16 unmarked vs. 9 marked), pune ‘put’ (24 unmarked vs. 16 marked), prinde ‘catch’ (12 unmarked vs. 3 marked), goni ‘chase’ (18 unmarked vs. 13 marked).

Notably, the language register used in these texts matches the language register of other translations of religious texts of the 16th centurry (ITZ, CC) and even of BB whose copy dates from the 17th century. This is important insofar as constructions found in these texts, and presented in Section 4 and Section 5 above, indicated that pre was used with direct objects not necessarily as a DOM marker but also as a case marker for accusative.

In addition to the examples in (12) and (14), there are other indications that the association of pre with a [case] feature was still at work in the 16th century: (i) the use of pre with inanimate direct objects; and (ii) the preponderance of pre with the fronted direct objects; see (21)–(23).

(i)

the use of pre with inanimate direct objects

(21)	a.	Şi	când	Domnezeu	pierdea	[pre	aceale	oraşă] (PO 82-83)
		and	when	God	lose.the	pre	that.pl	town.pl
		‘and when God was destroying these towns.’
	b.	ce	toiagul	lu	Aron	înghiţi		[pre toiagele		lor].
		and	cane.the	gen.m.sg	Aron	swallow.pst3sg		pre canes.the		their
		‘and Aron’s cane swallowed their canes.’				(PO 439)
	c.	Şi	găteaşte	[pre	mâncare.]		(PO 299)
		and	cook.3sg	pre	food
		‘and he cooks the food.’
	d.	nici blăstemă		[pre zioa]	ceaia ce-au născut		(EV 158)
		nor cursed		pre day.the	that that-has delivered
		‘nor did he curse the day in which he was born’
	e.	nu-ş	[pre niminea]	prepunea	el, numai	[pre păcatele		lui]	(EV 57)
		not-refl	pre nobody	blamed.3sg	he only	pre sins.the		his
		‘he wasn’t blaming anybody, only his sins’

In (21), the presence of pre cannot be attributed to DOM because the main priming factor for DOM is animacy, which is lacking with these direct objects. One may object that Table 1 does allow for DOM extension to inanimates, so (21) may signal just such an extension. However, Table 1 also entails that the extension of DOM to inanimates means that all the nominal categories that precede the inanimates on that scale must display obligatory DOM. That is clearly not what we see in the data of the 16th century. Hence, another explanation is in order for the possibility of pre with inanimates in (21). According to the discussion of the previous section, the most plausible alternative is to assume that the occurrences in (21) represent the tail-end of the analysis of pre as a case marker. This is a likely explanation, considering the conditions for the preservation of these texts: The versions of these texts that made it to the 20th century are the result of repeated copying by various monks, each copyst leaving his mark on the syntax of the text. Thus, it is likely that after the stabilization of the structural case assignment by verbs to their complements, pre has been deleted by copysts whose grammar had lost the analysis of this element as a case marker for direct objects. What we still find in texts could be either slips from previous versions or the reflection of the archaic language register of some copysts (the source texts for the translation do not have these objects marked). Be it as it may, the presence of pre as a case marker in religious texts conforms to the prescriptivism promoted at a later date in Eustativici’s grammar.

(ii)

the preponderance of pre with the fronted direct objects

Fronted objects undergo CLLD, which can be seen in both Old and Modern Romanian. However, the use of pe has changed: The examples in (18), from Modern Romanian, served to confirm the D (versus P) status of pe by showing it to be optional when the direct object is fronted above the verb. The optionality indicated that the fronted DP does not depend on pe for case. The situation is different in the 16th century texts, where pre is almost systematic upon CLLD. For example, in (22), the pronoun voi ‘you’ displays pre when fronted although its correlate mine ‘me’ in post-verbal position does not have pre.

(22)	De-au	gonit		[mine]	şi	[pre voi]	vor	goni; (EV 522)
	If-have.3pl		chased	me	also	pre you	will.3pl	chase
	‘If they chased me, they will also chase you’

The same can be observed for other categories, such as the quantifier ‘all’ in (23a) versus (23b), and common nouns and names (23c, 23d). That is, the fronted phrase displays pre irrespective of whether the same phrase in post-verbal position has or does not have pre.

(23)

a.

[pre tot oarece

era

în

oraşu

şi

în

câmp]

apucară

(PO 220)

pre all whatever

was

in

town

and

in

field

stole.3pl

‘whatever was in town and in the field, they looted it all’

b.

prădară

[tot oarece

era în

casele

lor]

(PO 119)

stole.3pl

all whatever

was in

houses.the

their

‘they looted whatever was in their houses’

c.

[pre Isus] cu multă

bună-govire  şi smerenie

ruga-l

(EV 367)

pre Jesus with much

good-respect and humility

beseached.3sg-cl.acc.3m.sg

‘as for Jesus, he beseeched him with much respect and humility’

d.

de

huliia

[fiiul

lu

Dumnezeu,

Domnul

nostru Isus Hristos]

(EV 198)

so.that

insulted.3sg

son.the

of

God

Lord.the

our

Jesus Christ

‘so that he insulted the son of God, our Lord Jesus Christ’

The fact that not only animate (23c, 23d) but also inanimate nominal phrases (23a) display pre when fronted suggests the following: in PO + EV, structural case assignment by verbs was stabilized for the direct objects in situ (hence, no need for a preposition), but not for the fronted ones (hence, a preposition is needed). That is, the structural case held for the post-verbal nouns but not for their copies when fronting applied. This condition would equally affect animate and inanimate nouns. However, it is also clear that animate nouns were preponderant in this configuration in which they are salient in the discourse and highly relevant to the information structure.

The relevance of animacy and discourse manipulation (i.e., CLLD) for the incidence of pre in the 16th century texts cannot be ignored, as these factors mingle with the ambiguity of pre as a case marker. In a nutshell, some other feature association is present for pre, beside the case requirement, since animacy entails DOM in a context in which these nouns receive a topic or a focus reading. The variation in the distribution of pre signals the weakening of the case function of pre (i.e., its P status) in the transition to a recategorization of pre as a nominal item (i.e., D) that serves as a DOM particle. Examples as in (24) support this transitional analysis.

(24)	Că şi	mai de nainte		tânjiia			ucenicii		de	aceasta,
	for even	more of before		coveted.3sg			apprentices.the		of	this
	unde	vedea	[Iacov]		şi	[pre Ioan]		cinstindu-i			(Ev 94)
	as	saw.3pl	Jacob		and	pre John		honoring-cl.3m.pl.acc
	‘Even before, the apprentices coveted this, as they saw Jacob and John honoring them’

There is asymmetric coordination in (24), which also appeared in (20e), confirming that the speaker treats a DP and a pre-DP in the same way; that is, these two structures qualify as syntactically equivalent, i.e., two DPs, which is the usual requirement for coordination. This brings further confirmation for the structure in (16).

Formally, this amounts to a modification of the features associated with pre as follows: [+/−case], [animacy], [d], which change the preposition pre into a DOM particle. The change consists in the loss of analyzability of pre as a case marker and the gradual replacement of [case] with [animacy] preponderantly in discourse contexts that entail topic interpretation (i.e., upon fronting), which also brings the discourse [d] feature in the syntactic computation of the DP. In 16th century texts, the feature bundle [animacy]/[d] is a stronger priming factor for DOM than [animacy] alone (i.e., more pre-DPs upon CLLD than in situ). The merging of pre in D versus P reflects this featural change, as the left periphery of DP was shown independently to encode discourse features (Giusti, 2005 for DPs in general; Irimia & Pineda, 2021 for DPs with pronouns under DOM).4

5.3. Animacy and Salience in the 17th Century

5.3.1. The Bible

BB was translated in the 17th century, but the use of pre with direct objects is different not only from the translations of the previous century but also from what can be seen in the other texts of the same century. The language register is archaic and yet influenced by changes that occurred in the language over a century. We treat this text as an intermediary type of grammar between PO + EV and the texts of the 17th century.

Remarkably, the use of pre with nouns and pronouns is almost systematic. Table 2 shows zero occurrences of unmarked animate direct objects with certain verbs (e.g., asculta ‘listen’; goni ‘chase’; omorî ‘murder’; ucide ‘kill’) and a reduced number of unmarked occurrences with other verbs. This indicates the stabilization of animacy as a priming factor for DOM at least with two nominal categories, i.e., pronouns and names. Also, pre is frequent with common nouns that denote not only humans but also animals, especially those assigned for religious sacrifice, as in (25a, 25b).

(25)	a.	Şi vei			aduce		[pre berbeace]			tot  pre jârtăvnic	(BB {Iesirea xxix})
		and will.2sg			bring		dom ram			also to scaffold
		‘and you will bring the ram also to the scaffold’
	b.	Roagă-te	dară cătră		Domnul		şi	să
		pray-refl	then   to		God		and	sbjv
		ia		de	la	noi	[pre şarpe]		(BB {NumerileCapXXI})
		take.sbjv.3sg		from	at	us	dom snake
		‘Pray to God and to take away from us the sneak’

When it comes to inanimate common nouns as direct objects, the use of pre went down considerably compared to the 16th century, although it still tends to appear with geographic names, as in (26a). Crucially, when the inanimate nouns are resumed by strong pronoun, as in (26b), pre is systematic, signalling the generalization of DOM to pronouns, irrespective of their reference.

(26)

a.

Şi

i-au

dat ei

Halev

[pre Golathmaim cea de sus]

and

cl.dat.3sg-has

given her.dat

Halev

pre Golathmain that of upper

şi

[pre Golath

cea

de

jos].

(BB {IisusNaviCapXV})

and

pre Golath

that

of

lower

‘and Halev gave her the upper Golathmain and the lower Golath’

b.

Şi

vei

face

jârtăvnic

de tămâiare

de

leamne

neputrede,

and

will.2sg

make

scaffold

for blessing

from

wood

not.rotten

şi-l

vei

face

[pre dânsul]

de un cot de lungu

(BB {IeşireaCapXXX})

and-cl.m.sg.acc will.2sg

make

dom it

of a elbow of length

‘and you wiil make a scaffold for the blessing from healthy wood, and you will make it one cubit in length’

There is still variation in the treatment of the same common noun when it comes to DOM, be it for humans or animals. Consider the minimal pairs in (27a, 27b) and (27c, 27d), each found in the same paragraph (hence, intra-speaker variation).

(27)	a.	voiu	îngropa	[pre mortul		mieu]	acolo		(BB {FacereaCapXXIII})
		will.1sg	burry	dom dead.the		my	there
		‘I will burry my deceased there’
	b.	şi voiu	îngropa	[mortul		mieu]	de la	mine	(BB {FacereaCapXXIII})
		and will.1sg	burry	dead.the		my	of	me
		‘I will burry my deaceased at my own expense’
	c.	Şi trimise	[porumbul]	după	dânsul	(BB {FacereaCapVIII})
		and sent.3sg	dove.def.m.sg	after	him
		‘and he sent the dove after him’
	d.	iarăş trimise	afară	[pre porumb]		den chivot
		again sent.3sg	outside	dom dove		from cage
		‘and he sent again the dove out of the cage’

These examples indicate that DOM responds to animacy on a systematic basis with pronouns and names of persons, but optionally with the rest of the nominal categories of the animacy scale. There is no evidence for an ambiguous treatment of pre as case marker or DOM particle in this text, the DOM function being well established.

Since animacy is not a systematic priming factor for DOM with common nouns, the question is what makes such a noun display pre at all. In BB, this can be related to the discourse context:

• Comparatives involving ca ‘as’

DOM in these comparative constructions was not systematic in the 16th century texts (see van Eeden, 1985; Stan, 2013), but becomes so in the 17th century. Compare (28a) from a 16th century text with (28b, 28c) from BB.

(28)

a.

Cinstiţi-i

ca

[nişte-ndreptători]

(van Eeden, 1985; YTZ 382)

honor.imp.2pl-cl.m.3pl.acc

as

some-justices

‘honor them as you would some court justices’

b.

mi-ai

loat

[featele

meale]

ca

[pre nişte roabe]

cu

sabia

cl.dat1.sg-have.2sg

taken

girls.the

my.pl

as

dom some slaves with sword.the

‘you took my girls by sword as if some slaves’ (BB {FacereaCapXXXI})

c.

Şi vei

farâma

[norodul acesta]

ca

[pre un om] (BB{NumerileCapXIV})

and will.2sg

break

people.the this

as

dom one man

‘and you will break these people as you would one man’

The direct objects in (28b, 28c) do not display pre, whereas their comparative correlates do.

• Salience in post-verbal position

DOM is resorted to when a new piece of information is introduced in the discourse and/or requires the spotlight in the narrative, as in (29a, 29c). The same nouns occur unmarked elsewhere where their reading is neutral (29b, 29d).

(29)

a.

văzu

[pre

un

om

eghiptinean]

bătând

[pre oarecare

saw.3sg

dom

a

man

Egyptian

beating

dom some

ovreaiu],

den

fraţii

lui

(BB {IeşireaCapI})

Jew

from

brothers.the

his

‘he saw an Egyptian beating some Jew belonging to his brothers

b.

ochii

lui văzu

[un om]

stând

(BB {IisusNaviCapV})

eyes.the

his saw.3sg

a   man

seating

‘his eyes saw a man seating’

c.

vindecă

Dumnezeu

[pre Avimeleh],

şi

[pre fămeaia lui],

heal.3sg

God

dom Avimelah

and

dom wife.the his

şi

[pre

slujnicele

lui]; {FacereaCapXX}

and

dom

servants.the

his

‘Good healed Avimeleh, and his wife, and his servants.’

d.

Sculându-te, ia

fămeaia

ta

şi

ceale doao feate     ale

waking-refl take.imp.2sg

wife.the

your    and   those two   daughters of

tale

ce

ai

şi

ieşi

afară, {FacereaCapXIX}

yours

that

have.2sg

and

go.imp.2sg

out

‘When you wake up, take your wife and those two daughters you have and go out’

Salience also entails enhanced referentiality, so DOM is also opted for in contexts with appositions (30a), repetitions (30b), information focus (30c).

(30)

a.

au

dat

Domnul

[pre  boiarii  lui   Madiam],

[pre   Oriv şi    pre  Ziv]

has

given

God

dom   nobles     of    Madiam

dom   Oriv and dom Ziv

‘God gave Madiam’s nobles, Oriv and Ziv

(BB {JudecătoriiCapVIII})

b.

au

luat

[pre boiarii]

şi

[pre cei

bătrâni

ai

cetăţii] (BB {JudecătoriiCapVIII})

have

taken

dom nobles.the and

dom those old

of

fort.the.gen

‘he took the nobles and the old men of the fort’

c.

Şi

făcu

Moisi [şarpe]

de aramă,{…]

Roagă-te

dară

and

made

Moisi snake

of copper

pray.imp.2sg-cl.2sg.acc

thus

cătră Domnul

şi

să

ia

de la noi

[pre şarpe] (BB {NumerileCapXXI})

to

God

and

sbjiv

takes from us

dom snake

‘Moisi made a snake from copper. Thus pray to God to take the snake from us’

• Salience in preverbal position (CLLD)

Fronting of the direct object for a topic or focus reading was identified as a priming factor for the use of pre in 16th century texts. In BB, most of the fronted objects display pre, although in their post-verbal position the same objects do not display pre. This is the case, for example, for fata ‘daughter’, which has no pre in post-verbal position in the entire text, but does so when fronted. Compare (31a) with (31b, 31c). However, instances of fronting without pre also occur with some noun phrases in the presence of quantifiers, as in (31d).

(31)

a.

[Pre fata

mea

aceasta]

am

dat

omului

acestuia fămeaie

dom girl.the

my

this

have.1

given

man.the.gen

this.dat woman

‘this daughter of mine I gave to this man as wife’

{LegeaIICapXXII}

b.

ca

să

iau

[fata

fratelui

stăpânului

mieu],

for

sbjv

take.1

girl.the

brother.the.gen

master.the.dat

my

‘so that I take the daughter of my master’s brother’

{FacereaCapXXIV}

c.

am

văzut

[fata

ta] {FacereaCapXXXIII}

have.1

seen

girl.the

your.f.sg

‘I’ve seen your daughter’

d.

Şi

[câte un boiariu

den fealiu]

veţ

lua  ca  să

vă

and

each one noble

from relation

will.2pl

take so sbjv

cl.acc.2pl

moştenească

de tot

voao

pământul. {NumerileCapXXXIV}

inherit

of all

to.you

land.the

‘And you’ll take a related noble to inherit your land’

In sum, BB attests to a stage where pre introducing direct objects is completely dissociated from [case] but strongly associated with [animacy] for certain nominal classes (i.e., pronouns and names). For these nominal items the discourse features are not relevant, since just animacy is sufficient to prompt the use of pre. For other nominal classes, however, [d] in addition to [animacy] is the trigger for the use of pre. In other words, DOM through the particle pre is well established in BB, but the priming factors are divided between [animacy] and [animacy]/[d].

5.3.2. Chronicles and Varlaam

The chronicles and Varlaam’s texts show a general elimination of pre with inanimate direct objects (with minimal exceptions). On the other hand, DOM increases with animate direct objects, as indicated in Table 2. For example, for the Ur text, 17 out of 25 verbs have zero occurrences of unmarked animate direct objects. Basically, in all three texts DOM became obligatory with pronouns and names, and the number of unmarked animate direct objects with other nominal categories is reduced when compared to BB. The verb trimite ‘send’ is an exception in Table 2, with a high number of unmarked objects. That is so because in the narratives, the persons sent with a message are indicated through common nouns (e.g., ‘messengers’, ‘envoys’, ‘soldiers’) which continue to allow for optional DOM with any verb, as shown in (32).

(32)

a.

au trimis

împotriva

lui

[slujitorii

săi] (Ur 109r)

has sent

against

him

servants.the

his

‘he sent his servants against him’

b.

au trimis

pre sluga

sa,

pre Crasiţchii la Liov,

(Ur 107v)

has sent

dom servant.the

his

dom Crasitsky to Liov

‘he sent his servant Crasitsky to Liov’

c.

umblându

pentru

domnie,

au lăsat

[nemţii]

şi

au trecut

la leşi (Ur 94v)

searching

for

throne

has left

Germans.the

and

has crossed

to Poles

‘in search for a throne, he left the Germans and crossed to the Poles’

d.

să

să

veselească

toţi

şi

cinstiră

[pre nemţi],

sbjv

cl.3pl.acc.refl enjoy.sbjv.3pl

all

and

honored

dom Germans

‘so all of them to enjoy themselves and honored the Germans’ (Ur 101v)

The contexts in which animate common nouns undergo DOM remain the same as in BB: comparative ca ‘as’ (33a), and salience for objects in situ (33b, 33c, 33d) or fronted (33e).

(33)

a.

toți

îl

iubiia

ca

[pre

un

parinte

(Ur 80v)

all

cl.3sg.acc.m

love.impf.3sg

as

dom

a.m.sg

parent.m.sg

al

său]

gen.def.m.sg

his

‘they all loved him as they would their parent’

- comparative

b.

au

avut

[doi ficiori],

[pre Stefan] și

[pe

Pătru]

(Ur 8v)

have.aux.3sg

had

two sons

dom Steven and

dom Pătru

‘he had two sons, Steven and Peter.’

- apposition

c.

au

tăiat

și

[pe

feciorii

lui

Arburie]

(Ur 66v)

have.aux.3sg

cut

also

dom

sons.the

gen.m.sg

Arburie

‘he also decapitated Arburie’s sons’

- contrastive focus

d.

au

bătut

[pre

o

samă de

oaste turcească]

(Ur 116r)

have.aux.3sg

beaten

dom

a.f.sg

some of

army Turkish.f.sg

‘he has beaten part of the Turkish army’

- information focus

e.

[pre maica-sa]

cu mare cinste

o

au

trimis

dom mother-his.f.sg

with high honor

cl.acc.f.sf

have.aux.3sg

sent

la domnu

său (Ur 42r)

- CLLD

to king.the

his.m.sg

‘the mother, he sent her to this king with great honors’

The importance of salience for DOM can be seen in asymmetric coordination where only one member of the coordination requires the spotlight and falls under DOM, whereas the other member remains unmarked, as in (34).

(34)	au	trimis	Dispot [nemții	săi]	și	[pre o samă de moldoveni] (Ur 100v)
	have.aux.3sg	sent	Despot Germans.the	his	and	dom a some of Moldavians
	‘Despot has sent his Germans and a group of Moldavians’

The construction in (34) is similar to the asymmetric coordination seen in PO + Ev (ex. 20e), indicating that a hundred years later DOM still maintains its optionality with common nouns, despite its stabilization with pronouns and names.

However, when it comes to animate common nouns, DOM is excluded with those denoting animals, as in (35a, 35b), unlike what happens in BB. In fact, such nouns do not display pre even when they undergo CLLD, as in (35c). There is, thus, a reduction in the animacy range for DOM.

(35)	a.	au		ucis		[acel	bour]		(Ur 8v)
		have.aux.3sg		killed		that.dom	aurochs
		‘he has killed that aurochs’
	b.	au		slăbitu		[caii	turcilor		cei gingaşi]	(Ur 26v)
		have.aux.3sg		weakened		horses.the	Turks.the.gen		those cute
		‘he weakened the Turks’ cute horses’
	c.	şi	[caii		turcilor]		au	slăbitu, (Ur 31r)
		and	horses.the		Turk.the.gen		have	weakened
		‘and he weakened the Turks’ horses’

In sum, the Chronicles and Varlaam’s texts of the 17th century attest to the development of DOM under animacy as the main priming factor in a way that limits animacy to humanness (a tendency that is preserved in Modern Romanian). This development conforms to the animacy scale in Table 2 by showing complete stabilization of DOM for the highest nominal categories, and optional DOM for the subsequent categories. For the latter, the option for DOM further involves discourse priming, so that the topic accessibility scale of Table 2 is also relevant to this process. In particular, CLLD (i.e., fronting to topic/focus) was shown to be the earliest context (activation of the topicality scale) in which pre became systematic with preverbal direct objects, and which favored an ambiguous analysis of pre as case marker and/or DOM particle, with the latter analysis expanding during the century that separated PO-EV from the Chronicles.

One scale of Table 1 that has not been relevant to the discussion so far is the definiteness/specificity. Indeed, this feature was orthogonal to the option for DOM in Old Romanian, as could be noticed in the examples provided so far for constructions with or without DOM. For example, the direct objects in (36), display the definite article and have a specific reading, yet DOM does not apply in any of the three texts.

(36)	a.	văzând Israil [feciorii		lui	Iosif]		(BB {FacereaCapXLVIII})
		seeing Israil sons.the		gen.m.sg	Joseph
		‘Israel seeing Joseph’s sons’
	b.	Din săraci şi		de rudă proastă			şi necărtulari	alease Hristos	[apostolii	săi] (Var 221r)
		from poor and		of family bad			and unread	chose Christ	apostles.the	his
		‘Christ chose his apostles from the ranks of the poor, of the low-life families and of the uneducated’
	c.	i-au	luat	dimpreuna si			[doamna]	si	[fiica	sa]	(Ur 25v)
		cl.dat.3sg-has	taken	together		and	wife.the	and	daughter.the	his
		‘he has taken his wife and his daughter together’

The insensitivity of DOM to the definiteness/specificity scale decreases towards Modern Romanian at the same time as the sensitivity to the topic accessibility scale also decreases. The context for this change is discussed in the next section.

6. Modern Romanian

This section presents the current results of DOM development in Romanian. The main observation is that DOM ceases to respond to salience but displays sensitivity to specificity and enhanced referentiality. That is, Romanian DOM changes the relevance of scales from topic accessibility to definiteness/specificity, showing loss of analyzability of pre as [d]. This change is related to the integration of clitic doubling in the pre-DOM construction, a process that was incipient in Old Romanian, as indicated in Table 2, but became predominant in Modern Romanian. The Modern Romanian data rely on current official grammars (GALR, 2008; GBLR, 2010; GR, 2013), current studies on Romanian DOM (Cornilescu, 2000; Mardale, 2009; Irimia, 2020; Tigău, 2020, a.o.) and the authors’ native speaker grammaticality judgments.

6.1. Clitic Doubling

Romance languages developed clitic pronouns out of Latin demonstrative and personal pronouns (Crysmann & Luís, 2024, a.o.), and Romanian makes no exception. Old Romanian displays clitic pronouns as substitutes to noun phrases (37a), as resumptive in constructions with CLLD (37b) and less frequently, as doubling the noun phrases in direct object position (37c).

(37)

a.

cu dobândă

va

să

ni-l

ceară

(BB {PrefaţăXXI})

with interest

will.3sg

sbjv

cl.dat.pl1-cl.acc.3sg.m

ask.sbjv.3sg

‘he’s going to ask it from us with interest’

b.

pre

fraţii

lui nu

i-au

cunoscut

(BB {LegeaIICapXXXIII})

dom

brothers.the

his neg

cl.acc.3pl.m -has

known

‘he did not recognized his brothers’

c.

să nu

le

aduceţ

aceastea

Domnului

(BB PreotiaXXII)

sbjv  neg

cl.acc.3pl.f

bring.2pl

these

God.the.dat

‘don’t bring these to God’

Clitic doubling (CD) as in (37c) and above in (2a) is the relevant configuration for the foregoing discussion since it represents a DOM strategy that does not involve a prenominal particle. More precisely, CD serves for DOM in Balkan languages and responds to definiteness/specificity, not to animacy (Hill & Mardale, 2021). The Balkan CD pattern is attested in Old Romanian in 16th century texts and in other religious translations such as BB cited in (37), but was unproductive and has disappeared by the 17th century. Crucially, although CD by itself, as in (37c), went out of use, it began to be integrated into DOM constructions with pre, as in (38).

(38)	i-au	mântuit	[pre ei]	Domnul	(BB {JudecătoriiCapII})
	cl.acc.3m.pl-has	absolved	dom them	God.the
	‘God has absolved them’

The integration of CD with pre-DOM is measured in Table 2, which shows a relatively small number of tokens in the 17th century. However, this option became productive toward Modern Romanian (Hill & Mardale, 2021) to the extent that in the 21st century it became obligatory with pronouns, names and common nouns that undergo DOM.

6.2. Specificity and/or Definiteness as DOM Triggers

Considering the DOM criteria in Table 1, CD and pre-DOM contrast w.r.t the scales they conform to; that is, the definiteness/specificity scale for CD versus the animacy scale for pre-DOM. The integration of the two strategies entails a change in the priming factors by adding specificity to the animacy. This scale combination is firmly stabilized in the 20th century, as the illustrations below attest by contrasting the grammaticality requirements for DOM constructions in Old and Modern Romanian.

• Demonstratives and possessives

In Modern Romanian, animate nouns preceded by demonstratives or possessives, which are inherently specific, must be marked through both CD and pe-DOM in colloquial register or even without CD in a more archaic register (hence, the optionality brackets around clitics).

(39)	Am	invitat(-o)	[*(pe)	această	fată]//	[*(pe) fata	lui].
	have.1	invited.cl.acc.f.sg	dom	this.f.sg	girl	dom girl.the	his
	Intended: ‘I’ve invited this girl//his daughter.’

DOM was not obligatory in similar contexts in Old Romanian in the absence of CD. Hence, the contrast between (39) and (40).

(40)	a.	cine	va	cinsti	[acesta]		(Ev 226)
		who	will.3sg	honour	this.m.sg
		‘who will honour this one’
	b.	n-au	luat	om [muiarea		lui]	la războiu	(BB {JudecătoriiCapXXI})
		neg-aux.3sg	taken	man wife.the		his	to war
		‘no man has taken his wife to the war’

The main difference between (39) and (40) is that the former displays CD, with inherent specificity reading, whereas the latter does not. Hence, diachronically, once CD became productive in conjunction with pe-DOM, the definiteness/specificity requirement for DOM was generalized.

• Complex noun phrases

Complex noun phrases increase the degree of specificity through the contribution of various modifiers. This was not sufficient to introduce the specificity trigger for DOM in Old Romanian (41a), while the addition of CD to this configuration did so in Modern Romanian (41b).

(41)	a.	chiemă înlăuntru	[toţi vrăjitorii	Eghipetului]
		call.3sg inside	all magicians.the	Egypt.gen
		‘he calls all the magicians of Egypt inside’
	b.	îi	cheamă înăuntru		[pe toţi vrăjitorii	Egiptului]
		cl.acc.m.pl	call.3sg inside		dom all magicians.the	Egypt.gen
		‘he calls all the magicians of Egypt inside’

• Ellipsis

In Modern Romanian, various nominal ellipsis contexts require DOM regardless of animacy; for example: the elliptical demonstrative (42a); the elliptical genitive (42b); some types of strong quantifiers (42c). DOM here can respond only to specificity, not necessarily to animacy.

(42)	a.	Îl	vreau	[*(pe)			acesta].
		cl.acc.m.sg	want.3sg	dom			this.m.sg.aug
		‘I want this one.’ (animate or inanimate)
	b.	Îl	vede		[*(pe)	al	tău.]
		cl.acc.m.sg	see.3sg		dom	gen.m.sg	your.sg
		‘He sees yours.’ (animate or inanimate)
	c.	Le	văd		[*(pe)	toate.]
		cl.acc.f.pl	see.1		dom	all.f.pl
		‘I see them all.’ (animate or inanimate)

Similar contexts display no DOM in Old Romanian, as shown in (43), unless the reference was to an animate entity and contextual salience was intended.

(43)

a.

au

auzit Domnul

cum mă

urăscu,

şi

have.aux.3pl

heard God

how cl.1sg.acc

hate.3pl

and

mi-au

dat

încă şi

[acesta]

cl.1sg.dat- have.aux.3pl

given

even also

this.m.sg.aug

‘God has heard how they hate me and he gave me also this on top’

(BB {FacereaCapXXIX})

b.

Să

nu

răpim

[a

striinilor],

nici

[al

său]

(EV 346)

sbjv

neg

steal.1pl

gen.f.sg

strangers.the.gen

nor

gen.m.sg

his

‘we should not steal what belongs to strangers or to our kin’

c.

eu

în locul

lui săvârşi-voiu

[toate],

EV 571

I

in place.the

his implement-will.1

all.f.pl

‘I will implement them all instead of him’

In sum, DOM sensitivity to definiteness/specificity coincides with the presence of CD in the derivation. Hence, CD introduced the definiteness/specificity as a priming factor for DOM, in addition to animacy. An important observation is that the productivity of the CD+pe-DOM combination led to grammaticalization in the sense that DOM applied irrespective of the discourse context, as a condition of grammaticality rather than a discursive option. That is, salience became irrelevant for the application of DOM, as this process became compulsory for pronouns, names, common nouns in complex phrases and various modifiers to elided nouns with specific interpretation. Instead of a salient reading, such nominal phrases bring referetial content that generally resumes a noun phrase previously introduced in the discourse (serving for discourse continuity in terms of Chiriacescu & von Heusinger, 2010, instead of discourse saliency).

Table 2 indicates that the combination of CD and pre-DOM is attested since the first preserved texts (i.e., PO and EV here), but its incidence was generally low over the entire period of Old Romanian. There are slightly more tokens of CD+pre-DOM in PO and EV than in the texts of the 17th century. This could be expected considering that PO and EV also attest more use of CD alone compared to the other texts. So there was ambiguity in the DOM options at the same time as the pre-DOM option was also emerging out of the loss of pre as a case assigner to selected noun phrases. However, once pre-DOM has stabilized in the 17th century, the CD+pre-DOM combination became unproductive, and was resuscitated later (i.e., starting in the 19th century; Hill & Mardale, 2021). In sum, considering the discourse effects identiffied for DOM in the 16th and the 17th centuries, pre-DOM emerged as a contrastive option to the unproductive CD: pre-DOM was foregrounding whereas CD was backgrounding the direct object (by resuming the referentiality of preceding items previously introduced in the discourse). The revival of CD in conjunction with pre-DOM must have occurred when the salient force of pre began to wear out.

6.3. Formalization

Clitics come in the derivation with inherent features preserved from the Latin etymon. More precisely, the Romanian clitics arise from the Latin pronominal series of ille, whose semantic feature composition is identified in Ledgeway (2012) as [+definite, -particularize, +given]. Accordingly, Romanian clitics systematically introduce definiteness and giveness, in addition to the formal features they respond to (i.e., object agreement) and irrespective of the syntactic function they fulfil (i.e., substitutes, resumptives, doubling). Hence, their interpretive effect is backgrounding and familiarity, and no formal discourse feature is needed in syntax for such a reading to arise.

This contrasts with the formal (vs inherent) [animacy] and [d] features of pre which, after the stabilization of DOM in the 17th century, catered to salient readings. As long as the association between [d] and pre was prevailing in the speakers’ grammar, there was no much room for a switch to DOM with CD. However, the main priming factor for pre-DOM was animacy, whereas [d] was secondary, being relevant only to the lower nominal categories of the animacy scale. As the productivity of pre-DOM increased and spread to more discourse contexts, the readings became ambiguous between saliency and neutral, and the association between [d] and pre weakened, that is, [d] became underspecified.

For example, the text in (44) taken from a 19th century narrative displays two instances of DOM. While the first DOM occurrence can be justified insofar as it situates the constituent in the spotlight, and hence, the salient interpretation, the second DOM occurrence has no discourse justification, as the constituent would receive a similar interpretation if the unmarked counterpart was used in that particular paragraph.

(44)

să

fi văzut

[pe neobositul

părinte]

cum

umbla

sbjv

be seen

dom relentless.the

priest

as

walked.impf.3sg

prin

sat

din

casă în casă,

împreună

cu bădiţa Vasile a Ilioaei,

through

village

from

house to house

together

with uncle Vasile of Ilioaia

dascălul

bisericii,

un holtei zdravăn, frumos

şi

voinic,

usher.the

church.gen

a bachelor healthy handsome and

sturdy

şi

sfătuia

[pe oameni] să-şi

dea

copiii

la învăţătură.

and

advised

dom people sbjv-cl.refl.3pl

send.sbjv.3sg children.the

to school

‘you should have seen our relentless priest going through the village from house to house, together with the church usher, Uncle Vasile of Ilioaia (mother’s name), who was a healthy, handsome and sturdy bachelor, and advised the people to send their children to school.’

(IC I)

Generally, a remarkable change appearing in IC’s narrative is that some nominal classes which displayed optional DOM in 17th century texts have obligatory DOM here; e.g., nouns denoting family relations (‘mother’, ‘father’, ‘daughter’, ‘aunt’) and quantifiers with or without the modified noun (‘each’, ‘all’, ‘many’). This extension is seen irrespective of the discourse context, which points to the dissociation of [d] from pre, as the salient value of [d] becomes diluted and ambiguous.

The incidence of CD is still low relative to the pe-DOM alone in this text (i.e., 33 pe-DOM vs. 11 CD+pe-DOM), but much higher compared to the CD + DOM in the 17th century texts in Table 2. Within the CD + DOM group, there are 5 quantifiers, 4 common nouns, 1 personal pronoun and 1 possessive ellipsis. This distribution suggests that CD affected first the nominal categories that are lower on the animacy scale and also low in referential content. The clitic contributes specificity to these categories by relating them to nouns previously introduced in the discourse.

From a syntactic perspective, the conjunction of CD and pe-DOM entails the configuration in (45).

(45)

In (45), pe remains associated with animacy but its underspecified [d] feature obtains its value from the local Spec-head configuration with the clitic; hence, pe becomes dependent on the definiteness/specificity of the clitic. Since ClP merges in the clause derivation in response to the object agreement feature of the TAM heads, the clitic undergoes movement and surfaces higher than pe-DP in the linear order (i.e., CD = Cl > pe-DP).

In sum, in Modern Romanian, the feature composition of pe in DOM constructions underwent a slight but consequential change insofar as discourse salience ceased to be a priming factor for DOM and the [d] feature became weak (underspecified) and/or completely dissociated from pe. The gap was filled by CD, which introduced a new priming factor for DOM, that is, definiteness/specificity.

7. Conclusions

The data presented in this paper indicate that, diachronically, pre has been associated with and dissociated from a series of formal features when it preceded a selected DP. More precisely, in this syntactic context, the 16th century texts provide evidence for the an ambiguous status of pre as case marker or incipient DOM marker catering to animacy. This is the stage in which the [case] feature is dissociated from pre as the [animacy] feature gradually replaces it. In 17th (and 18th) century texts, the analysis of pre as DOM particle is stabilized, the feature bundle associated with it is [animacy] and [d], and the priming factors can be identified as animacy and various types of discourse salience. In the background of this development of DOM persists a CD possibility, at an unproductive level. This option surfaces in the 19th century when the [d] feature of pe is worn out, and DOM extends as compulsory to more nominal categories than in the previous century. CD is revived in conjunction with pe and triggers a change in the priming factors for DOM, by introducing definiteness/specificity in addition to animacy, as the [d] feature is gradually dissociated from pe.5

The loss of analyzability approach allowed us to capture the gradual change in the feature composition of pre as a DOM particle, since the texts display intra- and inter-speaker variations that indicate concurrent different analyses of this element during centuries. The context of use provided important clues for the path of DOM development, e.g., it indicated a predominant option for pre-DP under CLLD/topic reading in the emerging stage, followed by the pairing of pre-DP with the placement of the object in the spotlight even in post-verbal positions. From this perspective, the preservation of the feature composition of pe with adjunct DPs but not with the selected ones becomes predictable as the two syntactic environments cater to different contexts of use.

Corpus
BB	Chiţimia, Ion C (ed). 1988. Biblia, adecă Dumnezeiasca Scriptură (1688). Bucureşti: Editura Institutului Biblic. (www.sfantascriptura.com)-format pdf Re-edited in 1988. Bucureşti: Editura Institutului Biblic şi de Misiune Ortodoxă.
CM	Dragomir, Otilia (ed). 2006. Constantin Cantacuzino, Istoriia Ţărâi Rumâneşti. Bucureşti: Editura Academiei Române.
DÎ	Chivu, Gheorghe et al. (eds). 1979. Documente şi însemnări româneşti din secolul al XVI-lea. Bucureşti: Editura Academiei Române.
DRH	Pascu, Ştefan et al. (eds). 2003. Documenta Romaniae Historica. (17th-19th c.). Bucureşti: Editura Academiei Române.
E	Ursu, Nicolae A (ed). 1969. D. Eustatievici Braşoveanul. Gramatica rumânească (1757). Bucureşti: Editura Ştiinţifică.
EV	Puşcariu, Sextil & Procopovici, Alexie (eds). 1914. Carte cu învăţătură (1581). Bucureşti: Atelierele Grafice Socec & Co.
IC	Creangă, Ion.1892 [2010]. Amintiri din copilărie. Bucureşti: Editura Gramar.
ITZ	Eeden, van W. (ed). 1985. Învăţături preste toate zilele (1642). Amsterdam: Rodopi
PO	Pamfil, Viorica (ed). 1968. Palia de la Orăştie (1581–1582). Bucureşti: Editura Academiei Române.
Ur	Panaitescu, Petre P (ed). 1958. Grigore Ureche, Letopiseţul Ţării Moldovei (1642–1647). Bucureşti: Editura de Stat.
Var	Byck, Jacques (ed). 1964. Varlaam. Cazania (1643). Bucureşti: Editura Academiei Române.