Constructing Meaning: Historical Changes in mihi est and habeo Constructions in Romanian

Abstract

In this article, I address the evolution of the competition between two Latin patterns, habeo and mihi est, in Romanian. As opposed to the other Romance languages, which replace the mihi est pattern with habeo in possessor and experiencer contexts, Romanian maintains both Latin patterns. The general evolution of these patterns in the Romance languages is well known, however, a detailed usage-based account is currently lacking. Building on the theoretical findings on the role of functional competition in linguistic change, the rivalry between the two patterns in Romanian has already been argued to have settled in terms of differentiation, with each of the two forms specializing in different functional domains by Vangaever and Ilioaia in 2021 in their study “Specialisation through competition: habeo vs. mihi est from Latin to Romanian“. With this idea as a starting point, I investigate, by means of a diachronic corpus study, whether the dynamics in the inventory of state nouns occurring in these constructions can affect their evolution and productivity. The preliminary results show that this is indeed the case. Concomitantly, I explore whether the historical changes that the two patterns have undergone over the centuries can be described in terms of grammaticalization, constructionalization, or in terms of constructional change.

1. Introduction

This paper deals with the two Romanian constructions illustrated in (1) and (2) below. The construction in (1) is an instantiation of a mihi est construction (cf. Ilioaia 2021; Ilioaia and Van Peteghem 2021), in which the verb fi ‘be’ occurs with a noun denoting a state (henceforth ‘state noun’) and a dative experiencer, a combination that conveys to the construction the meaning of a psychological or a physiological state. The construction in (2) illustrates the habeo construction and expresses exactly the same meaning, this time from the combination of the verb avea ‘have’ with a state noun and an experiencer in the nominative.

(1)	Băiatului	îi	e	frică	de	poliție
	boy	him.dat	is	fear	of	police
	‘The boy is afraid of police.’
(2)	Io	tot	am	frică	la	dentist
	I	still	have	fear.acc=nom	at	dentist
	‘I am still afraid when I go to the dentist.’						(ioanaspune.ro, accessed on 5 May 2018)

In Romanian, the mihi est, as well as the habeo constructions, developed from the inherited Latin patterns (cf. Benveniste 1966, p. 197; Bauer 1996). They have both been preserved in Romanian, while, in other Romance languages, the habeo construction entirely replaces the mihi est construction. The two inherited constructions are in competition in Romanian (Ilioaia 2021; Vangaever and Ilioaia 2021; Ilioaia 2023), and they are argued to have already been in competition with each other in Latin (cf. Fedriani 2011 p. 312; Baldi and Nuti 2010, pp. 260–61). With respect to the historical changes that these two constructions underwent, I will explore whether they can be accounted for in terms of grammaticalization, constructionalization, or in terms of constructional change.

The general evolution of these patterns is well known; however, a detailed account is currently lacking. This may be due to the absence of a coherent theoretical framework to properly assess this evolution and the lack of quantitatively oriented corpus-based studies addressing the transition from Latin to the individual Romance languages.

I intend to fill this gap to a certain degree by addressing the peculiar evolution of the two patterns in Romanian. Therefore, in the present paper, I aim to answer the following research questions: (i) how did the two constructions evolve during the period starting with the first Romanian written text until the present day?; could this evolution be described as an instance of constructionalization, or do we witness a series of constructional changes happening within the same constructional node?; and (ii) how do changes in the inventory of the state nouns occurring in these constructions in Romanian affect the evolution and the productivity of the constructions themselves?

In order to provide an answer to these research questions, I will test by means of empirical studies the following hypotheses. With respect to the first research question, my hypothesis is that the evolution of the two constructions in Romanian represents a case of constructionalization, with the mihi est as a new construction conveying a psychological and a physiological meaning and the habeo construction becoming typical for abstract possession. As for the second research question, I hypothesize that the investigated state nouns first occurred in the habeo construction with an experiential meaning and have shifted over time to the experiential mihi est construction, which has become more appropriate for expressing physiological and psychological states with a temporary nuance. Such an evolution also generates the expectation that these specific state nouns occurred more frequently with the habeo experiencer construction in old Romanian than in the present-day language.

2. Theoretical Approach

In order to answer the research questions stated above, I adopt a constructionist, usage-based approach, in which I combine elements of the Construction Grammar framework (Langacker 1987, 2003; Fillmore 1996; Croft 2000; Croft and Cruse 2004; Goldberg 2006; Hoffmann 2013), with elements pertaining to the Diachronic Construction Grammar framework (Fried 2009; Traugott and Trousdale 2013; Barðdal et al. 2015). As for the competition between the two constructions in Romanian, I will make use of a model of language change based on the concept of functional competition (De Smet et al. 2018, p. 198).

2.1. Construction Grammar

The two constructions under scrutiny are conceived of as constructions in the sense described in Construction Grammar. In this framework (cf. in particular Goldberg 1995, 2006, 2013; Croft 2001), constructions are defined as conventionalized associations of form and function: [[Form] ↔ [Meaning]]. Such associations are unique and conventionalized and show different degrees of internal complexity and lexical specificity. In addition to words (e.g., book), they include, among other things, prefixes (e.g., pre-) and idioms (e.g., get one’s ducks in a row).

As opposed to the modular approach of Generative Grammar, in which the arbitrary character of language is entrusted to the lexicon, while syntax consists of the organization of lexical units, Construction Grammar also acknowledges syntactic constructions to be pairings of form and function (e.g., John gives/slides the book to Mary). For example, the ditransitive construction involves a subject, a verb, and two objects on the formal side and evokes a transfer of an entity by an agent to a beneficiary on the functional side (Goldberg 2006, p. 5). Due to the association of this meaning with the form, it is possible for verbs like slide, which do not intrinsically evoke a transfer, to express, nonetheless, such a scenario, and thus to behave like give, a verb denoting a transfer through its lexical semantics (Goldberg 2006, p. 7). Constructions are seen as the minimal units of language. Their inventory in a given language is referred to as the “constructicon”, by analogy to the term “lexicon” (Goldberg 2019, pp. 35–37).

2.2. Grammaticalization, Constructionalization, and Constructional Change

Originally, grammaticalization denoted a process through which a lexical item changes over time into a morphological item (Meillet 1912). Although it is a relatively old concept with a long history, especially in the field of morphology (Meillet 1912), the concept of grammaticalization has gained an “expanded” meaning during the last half of the century, being mentioned in nearly all of the studies involving language change and identified with the processes that create such changes, as well as with the theory modeling these changes itself (cf., among others, Kuryłowicz 1965, p. 69; Lehmann [1982] 2015, 2002, p. 7; Burling 1992, p. 300; Traugott and Trousdale 2013; Hilpert 2013, 2018; Barðdal and Gildea 2015; Gildea and Barðdal 2023). This controversial “expanded” concept of grammaticalization has more recently been linked with the framework of Diachronic Construction Grammar, which is seen as its “well-elaborated agreed-upon theory” (Gildea and Barðdal 2023, p. 2; Noël 2007; Gisborne and Patten 2011; Hilpert 2013, 2018; Coussé et al. 2018). Indeed, terms such as lexicalization, constructionalization, or even pragmaticalization are often defined by invoking a specific flavor of the well-known concept of grammaticalization.

Constructionalization is a more recent concept. It was first used by Rostila (2004) and then by Noël (2007). It is used to refer to “the development through which certain structural patterns acquire their own meanings, so that they add meaning to the lexical elements occurring in them” (Noël 2007, p. 192), as is the case with the way construction (The wounded soldiers limped their way across the field, Noël 2007, p. 187). Traugott and Trousdale (2013, p. 22) are stricter in defining this concept. For them, constructionalization is defined as the creation of a form−meaning pairing representing a new type node with its own syntax or morphology and also its own (new) coded meanings, a pairing that has been replicated across a network of language users. Everything that happens with the construction before and after the creation of this new node in the network (before and after the constructionalization) is described as effects of pre-constructionalization and post-constructionalization, which are basically constructional changes.

Constructional changes, as defined by the same scholars (Traugott and Trousdale 2013, p. 22), represent changes affecting one internal dimension of a construction, and they do not involve the creation of a new node in the network. Importantly, these changes in contextual use can be observed prior to and following the constructionalization of a pattern (cf. Traugott and Trousdale 2013, pp. 22–26; Hilpert 2015, pp. 135–36).

2.3. Functional Competition

To account for the competition between the habeo and the mihi est constructions, especially after they have been passed on from Latin to Romanian, I will make use of the concept of functional competition. In linguistics, the concept of functional competition is a commonly used metaphor to refer to the “selection struggle” among alternative forms encountered by the speaker during language production (Fonteyn 2019, p. 53; Berg 2014, p. 344). The strength of this struggle depends on the degree of formal or functional similarity between the alternatives, as follows: the higher their similarity, the stronger their competition (Berg 2014, p. 344). Competition is the strongest when the alternatives combine formal and functional similarities (Berg 2014, p. 344).

Once two forms enter into competition with one another over some functional domain, two main scenarios can be distinguished (cf. De Smet et al. 2018, p. 198). Either one of the alternatives is favored, causing its competitor to decline and even disappear (Leech 2009), or each of them subsists but specializes in distinct functional domains (Cacoullos and Walker 2009). The outcomes of these scenarios, called respectively substitution and differentiation by De Smet et al. (2018, p. 198), have a similar effect on the language system: they increase its degree of isomorphism (De Smet et al. 2018, pp. 198–99). Isomorphism refers to the situation in which there is a functionally motivated division of labor between the forms of a given language, in line with the more general idea that “a difference in syntactic form always spells out a difference in meaning” (Bolinger 1968, p. 127). Languages are claimed to naturally develop an increasing degree of isomorphism, ideally providing their users with one form for one function (al domain).

In the present paper, I will look into the possibility that (i) the two experiencer constructions under scrutiny may have evolved in Romanian into a new node in the constructional network, hence they may have constructionalized; or (ii) that they may instantiate cases of constructional change, limited to transformations affecting one internal dimension of the construction. The empirical data analyzed in the present study also aim to provide additional evidence in favor of the claim put forward by Vangaever and Ilioaia (2021), namely that the path of the functional competition between the two constructions settles in Romanian in terms of differentiation and not in terms of substitution.

3. The Evolution of the mihi est and habeo Patterns

3.1. The Journey of the mihi est and habeo Patterns from Latin to Romance

In Latin, prototypical possession is realized in several ways. Prototypical possession is defined as a relation of ownership between an animate possessor and a concrete possessee that can, in some way, be used by the possessor (Bolkestein 2001, p. 269). Among the several strategies in Latin, the following two cross-linguistically recurrent patterns can be distinguished: a transitive pattern, in which a verb of possession occurs with a (potentially pro-dropped) nominative NP and with an accusative one (3), namely the habeo pattern; and an intransitive pattern, in which the verb esse ‘to be’ is combined with a nominative NP and a dative one (4), more specifically, the mihi est pattern (Bolkestein 2001, p. 269).

(3)	Librum	habeo				(Possessor habeo)
	book.acc	have
	‘I have a book.’
(4)	Mihi	est	liber			(Possessor mihi est)
	I.dat	is	book.nom
	‘I have a book.’

In these structures, the nominative NP in the transitive pattern and the dative NP in the intransitive pattern encode the possessor, while the accusative and the nominative NPs, respectively, encode the possessee. Although the possessee is mostly concrete in Latin, it may sometimes be abstract, as in (5) and (6), respectively.

(5)	Febrim	habeo				(Experiencer habeo)
	fever.acc	have
	‘I have a fever.’
(6)	Mihi	febris	est			(Experiencer mihi est)
	I.dat	fever.nom	is
	‘I have a fever.’

These abstract NPs encode the possessee, and hence the transitive and intransitive patterns in (5) and (6) do not express a relation of prototypical possession, but an experience (Bolkestein 1983, pp. 83–84; 2001, p. 269; Fedriani 2011, p. 310; Pinkster 2015, p. 108; Danesi and Barðdal 2018, p. 23). The two patterns may be labeled possessor habeo and possessor mihi est when they occur with a concrete noun, as in (3) and (4) above, and experiencer habeo and experiencer mihi est when they occur with an abstract noun, as in (5) and (6).

The examples above illustrate that possessor habeo (3) and mihi est (4) have distinct formal, but similar, functional properties and that the same holds for experiencer habeo (5) and mihi est (6). Since constructions with similar functional properties often enter into competition, it comes as no surprise that competition exists between habeo and mihi est in both possessor and experiencer constructions (Baldi and Nuti 2010, §2; Fedriani 2011, pp. 310–11). This competition exists in Latin from the earliest texts onward and might even be inherited from an earlier stage of the Indo-European language (Fedriani 2011, p. 311).

However, as shown by Vangaever and Ilioaia (2021), the competition between habeo and mihi est is not absolute (cf. also Fedriani 2011, pp. 310–11). More precisely, in Latin, a habeo construction can always be replaced by a mihi est construction, while the opposite substitution does not hold. This imbalance between habeo and mihi est patterns is argued to be due to a difference regarding the nouns that occur in possessor and experiencer constructions, as follows: the variety of the concrete nouns coding the possessee in possessor constructions is open with both habeo and mihi est, while the inventory of the abstract nouns filling the stimulus slot in experiencer constructions is restricted with both verbs, though more significantly with habeo (Baldi and Nuti 2010, pp. 260–61; Fedriani 2011, p. 311). This shows that mihi est is more specialized in experiencer contexts than habeo. In the oldest texts, e.g., those of Plautus, experiencer mihi est is, moreover, preferred over habeo, which is reflected in the higher frequency estimated by Baldi and Nuti (2010, pp. 260–61) at 37 experiencer mihi est vs. 7 experiencer habeo examples. However, instances of experiencer habeo have already been attested in the earliest period (Fedriani 2011, pp. 311–12), contrary to what is claimed by Löfstedt (1963, pp. 76–78). In spite of the lack of quantitatively oriented corpus studies for Classical Latin, it has been argued that, from the 1st century BC onward, experiencer uses of habeo, as shown in (7), gain in frequency, gradually putting an end to the absolute dominance of mihi est (Fedriani 2011, pp. 311–12; Löfstedt 1963).

(7)	Si	cui	venae	sic	moventur,
	if	rel.dat	veins.nom	so	tremble
	is	habet	febrim
	he	has	fever.acc
	‘If his veins tremble in this way, he has a fever.’						(Cic. De fato 15)

Löfstedt (1963; see also Fedriani 2011, pp. 310–11) argues that the increasing use of habeo instead of mihi est in experiencer constructions is due to a gradual loosening of the original constraints on the object of habeo, i.e., that it should be a concrete noun denoting an object that can somehow be used by the possessor. Hence, in Classical Latin, speakers came to gradually exploit the use of habeo in experiencer contexts, which led to a more intense selection struggle between this construction and experiencer mihi est.

According to Fedriani (2011, p. 310), the transition from concrete to abstract possession (or experience) can be analyzed within the context of the “ideas are objects metaphor.” This metaphor conceives of the mind as a container in which abstract entities, such as feelings and emotions, are stored as objects (Kövecses 2003, p. 89). This is in line with the more general idea that the use of abstract nouns consists of “a linguistic technique that allows actions and processes to be treated as if they were things” (Seiler 1983, p. 52).

In Late Latin, habeo and mihi est still occur in both possessor and experiencer constructions. At this stage of the language, however, the competition between habeo and mihi est increases in intensity, and, more significantly, starts settling in favor of habeo, both in possessor and in experiencer constructions, moving towards total replacement of mihi est by habeo in the Romance languages (Fedriani 2011, p. 311). Knowing that the two main outcomes of functional competition are substitution and differentiation, the generalization of habeo at the cost of mihi est should be analyzed, at this point, in terms of “constructional substitution” (Fedriani 2011, p. 311). The empirical evidence supporting this evolution comes from the Romance languages themselves. While mihi est is entirely lost in most of these languages, habeo can still take as its object both concrete (8a–c) and abstract (9a–c) nouns, and thus occurs in possessor, as well as experiencer, constructions (Stolz et al. 2008; Van Peteghem 2017).

(8)	a.	Pierre	a	un	livre			(French)
	b.	Pedro	tiene	un	libro			(Spanish)
	c.	Pietro	ha	un	libro			(Italian)
		Peter	has	a	book
		‘Peter has a book.’
(9)	a.	Pierre	a	faim				(French)
	b.	Pedro	tiene	hambre				(Spanish)
	c.	Pietro	ha	fame				(Italian)
		Peter	has	hunger
		‘Peter is hungry.’

Vangaever and Ilioaia (2021, p. 254) raise the question of why the competition between habeo and mihi est settled in the Romance languages in terms of substitution and not in terms of differentiation. One possible reason could be the perception, already existing in Latin, of habeo being a more expressive construction (Fedriani 2011, pp. 312–13). For instance, grammarian Donatus explicitly testifies to the expressivity of habeo in his comment on verse 40 of Terentius’ Andria, the 4th century AD, as follows: Plus dixit “in memoria habeo” quam si dixisset “scio” (‘“I have in memory” means more than “I know”’). As a second reason, the scholars invoke a more general tendency observed in the evolution of the Indo-European languages toward a more transitive syntax (cf. Bauer 1993, p. 65). The two scholars tend to consider the syntactic pressure exerted by the spread of transitivity as having a more significant role in this substitution. However, sincethe replacement of the intransitive mihi est pattern by the transitive habeo pattern may be considered as part of a large-scale Indo-European syntactic drift, one might raise the question of why it did not take place in the transition from Latin to Romanian.

3.2. Exploring the Evolution of the Competing Constructions mihi est and habeo in Romanian

Indeed, contrary to the other Romance languages, Romanian has preserved both habeo and mihi est, in possessor, as well as experiencer, contexts (10) and (11) (Vangaever and Ilioaia 2021; Niculescu 2013, pp. 185–86). It has to be noted that the use of the mihi est pattern in possessor contexts is restricted to identificational clauses (11a). The nominative NP is the predicate of another, potentially pro-dropped, nominative NP (e.g., Ioana, in this example) acting as the subject of fi ‘to be’ and being referentially identified by the “property-denoting nature of the possessee NP” (Niculescu 2013, p. 186).

(10)	a.	Am	un	fiu
		have	a	son.acc
		‘I have a son.’
	b.	Am	frică
		have	fear.acc
		‘I am afraid.’
(11)	a.	(Ioana)	Îmi	este	cumnată
		Ioana.nom	me.dat	is	sister-in-law.nom
		‘Ioana is my sister-in-law.’
	b.	Mi-	e	frică
		me.dat-	is	fear.nom
		‘I am afraid.’

The survival of habeo and mihi est in the two constructions excludes the possibility that their competition in the transition from Latin to Romanian settled in terms of substitution. Given that the other main outcome of functional competition is differentiation, it is thus expected that habeo and mihi est specialized in different possessor and experiencer contexts.

In their paper, Vangaever and Ilioaia (2021) investigate how this competition has evolved in Romanian from the beginning of the 16th century (with the first Romanian text dating from 1521) to the present-day language via a corpus study. Through this study, the two scholars verified the following two main hypotheses: (i) the Possessor Experiencer Differentiation Hypothesis and (ii) the Experiencer Differentiation Hypothesis. By invoking the Possessor Experiencer Differentiation Hypothesis, they assumed that the competition between the two constructions evolves toward a neat division of labor, in which habeo becomes restricted to possessor constructions and mihi est to experiencer constructions. By invoking the Experiencer Differentiation Hypothesis, the two scholars expected that habeo is preserved in experiencer constructions and that it specializes in contexts from which experiencer mihi est is excluded.

The results of the corpus study carried out by Vangaever and Ilioaia (2021) contradict the first hypothesis, while the second hypothesis is confirmed. More specifically, the examined data show that, in Romanian, habeo does not become restricted to possessor constructions, but occurs also in experiencer constructions, where it shares an increasing inventory of abstract nouns with experiencer mihi est. As for the second hypothesis, the data show that the competition between the two constructions, which takes place this time within the experiencer domain, settles in terms of a specialization in distinct sets of abstract nouns. Indeed, experiencer mihi est occurs in present-day Romanian increasingly with abstract nouns expressing psychological and physiological states. To the contrary, experiencer habeo, which, in Old Romanian, occurred with a larger inventory of nouns denoting psychological and physiological states, occurs in the present-day language with a diversity of abstract nouns denoting various states, but to a much lesser extent with nouns expressing psychological and physiological states. Nevertheless, as pointed out by a reviewer, habeo is still well-formed with psychological or physiological states in generic, presentational contexts, especially when there is further modification of the noun denoting a state, as, for example, in Ion are o foame de lup ‘Ion has a hunger of wolf’ (meaning that he tends to eat a lot), whereas the experiencer mihi est seems to rather convey a temporary nuance to the state expressed.

As is well known, the lexical variation in a specific slot of a syntactic construction is an important parameter to measure the productivity of a construction, i.e., its degree of schematicity and, from the point of view of the language user, its applicability. The more lexical variation there is, the higher the degree of schematicity it has, and, hence, the more productive the construction is (Barðdal 2008, p. 22). The lexical variation that characterizes the set of abstract nouns occurring in the experiencer habeo construction may thus be linked with a higher degree of productivity, as compared to the experiencer mihi est construction, which is limited to nouns expressing psychological and physiological states. In contrast, the low lexical variation (fewer types) of the set of nouns occurring with the experiencer mihi est, in combination with a high semantic coherence, indicates a lower degree of schematicity, and, consequently, a lower position on the productivity cline (Barðdal 2008, p. 38). The schematicity of a pattern is the result of the process of entrenchment, which is defined as the continuous reorganization of linguistic knowledge caused by the repeated usage activities in usage events (cf. Barðdal 2008). The more entrenched and schematic a pattern is, the higher the chance of it evolving into a new node in the network, i.e., as a new construction with its own meaning.

Indeed, Ilioaia (2020, 2021), Ilioaia and Van Peteghem (2021), and Niculescu (2013, p. 186) observed that the mihi est pattern is mostly associated with nouns from the field of psychological and physiological states in Romanian, while, for habeo, no semantic categories have been proposed. In line with the Experiencer Differentiation Hypothesis, Vangaever and Ilioaia (2021) show that, over time, the nouns used with habeo tend to be different from those occurring with mihi est. The empirical data on Romanian analyzed by the two scholars indeed show that the use of habeo with psychological and physiological stimuli gradually decreases from the 17th century until today, with such configurations being perceived as ill-formed in the present-day language (e.g., ? am frică ‘I’m afraid’).

Bearing this in mind, the question arises as to how the two constructions interact with the set of nouns with which they occur throughout the centuries. Does mihi est occur with nouns other than those selected by the habeo construction? And, if this is the case, can this evolution be described in terms of constructionalization? Does mihi est receive a degree of schematization in Romanian that allows it to be considered as a new node in the network of the construction, as opposed to the situation in Latin?

These issues will be tackled in the next sections, where I first investigate, from a diachronic perspective, the dynamics in the evolution of a select set of nouns occurring in the two constructions. Then, I will address the arguments in favor of an analysis in terms of constructionalization of the mihi est in Romanian.

4. Empirical Case Studies

4.1. Methodology

In order to understand the dynamics of the set of nouns that occur with the mihi est and habeo constructions, I carried out two corpus studies based on texts from pre-21st century Romanian and the present-day language. For pre-21st century Romanian, I worked with a corpus made by myself, which is accessible on demand for research purposes on the Sketch Engine platform. This corpus, labeled Pre-21st century Romanian, contains nearly six million words. As for the present-day language, I worked with the Romanian Web 2016 (roTenTen16) corpus containing over two billion words, which was compiled and made available on Sketch Engine. It has to be specified that the collection of data was a two-step process, as follows: the detection of the nouns occurring with the mihi est construction in Romanian was a necessary step in order to reveal all of the other experiencer structures in which these nouns occur, and to avoid, at the same time, an enormous amount of noise coming from the frequency and flexibility of the verb fi ‘be’ on the one hand, and that of the nouns that combine with it on the other hand.

Using the advanced query option on Sketch Engine, all examples containing the verb fi ‘be’, preceded by a dative clitic and followed by a noun, were gathered from the corpus for present-day Romanian, roTenTen16, provided by Sketch Engine, and the corpus for pre-21st century Romanian, created by myself. By means of an advanced query, I searched for all structures of the type [DAT fi N] (cf. mi-e lene ‘I am lazy’), where N stands for any noun that can combine with an experiencer in the dative and the verb fi ‘be’.

The query returned 154,492 examples for present-day Romanian, which I restricted to a random sample of 100,000 examples, the maximum amount of data that can be exported from Sketch Engine. The same query in the corpus for pre-21st century Romanian returned 2,278 examples, which were all preserved.

For each of the two samples, I automatically generated a list of different sequences (of the type mi-e foame ‘I am hungry’ or li-era foame ‘they were hungry’) using the Frequency tool on the Sketch Engine platform. The full list was then exported to Excel and manually annotated. After removing the noise, the remaining examples were further annotated, and a final list was made containing the nouns occurring in this construction in all periods of Romanian. Due to imperfections in tagging or inconsistency in the use of diacritics in the older texts, the query returned a larger amount of noise for pre-21st century Romanian than for present-day Romanian. In spite of that, the data for pre-21st century Romanian allowed me to collect 29 different state nouns occurring in the mihi est construction, whereas the corpus for the present-day language yielded 95 different nouns in this construction.

The aim of the second phase of the corpus study was to search for all structures that can host the gathered nouns, besides the mihi est construction, both in present-day Romanian and in pre-21st century Romanian. Several structures, such as the habeo construction, and other verb constructions with a nominative, accusative, or dative experiencer, were found. To achieve the maximum number of each of the selected structures for each noun, and to minimize the amount of noise, I created very specific queries for each of the nouns from the list collected during the first phase of the study. An example of a simplified query is [dat v n]|[nom v n]|[acc v n], where ‘v’ can be any verb, including fi ‘be’ or avea ‘have’, and n is each time replaced with one of the nouns from the list gathered in the first phase, with a distance of zero to three words between them.

Several peculiarities of the present-day language were considered, such as irregularities in the use of diacritics (cf. frică vs. frica, scârbă vs. scarbă/scârba/scarba/scîrbă/scîrba/scirbă/scirba, etc.), or spelling modifications for pragmatic reasons (cf. frig vs. friiiiig, or frică vs. fricăăăă/fffrică, etc.). As for the particularities of old Romanian, special attention was given to archaic forms (cf. hi in place of fi ‘be’, pohtă instead of poftă ‘craving’, seate for sete ‘thirst’, etc.) and to inconsistencies in the already mentioned use of diacritics (cf. supra). It goes without saying that, when working with old texts, other orthographic peculiarities may occur, which cannot always be predicted (cf. for instance, words broken by brackets). These issues can explain the greater amount of noise that was extracted from the corpus for pre-21st century Romanian.

After considering all of the predictable situations, I ran the query for each noun. From the total number of hits per noun, a sample of 200 sentences was selected, with my goal being to collect approximately 100 relevant examples per noun in order to generate the final dataset. It has to be noted that this was not possible for all of the nouns or for all periods, due to size-related limitations.

By means of these queries, I was able to retrieve 16,550 examples for all periods of Romanian, before eliminating the noise (i.e., not relevant contexts, doubles, or typesetting errors). These examples were extracted and saved in an Excel document. After manually filtering the noise, as well as the examples containing constructions other than habeo and mihi est, my sample for the present study contains 7,007 examples to be analyzed (tokens), among which 3,969 examples (57%) are instantiations of the mihi est construction, while 3,038 examples (43%) contain the habeo construction. As expected, the relevant examples from present-day Romanian are more numerous than those from pre-21st century Romanian. More precisely, 73% (5,094) of the examples represent present-day Romanian, and 27% (1,913) pre-21st century Romanian.

With respect to the gathered set of nouns occurring in the two constructions during both present-day and pre-21st century Romanian, the majority of them are old nouns, either inherited from Latin or loaned from Old Slavic. A few of them are eliminated from the mihi est construction in the 19th century (nevoie ‘need’) or the 21st century (jind ‘craving, longing’)—while they are still used in present-day Romanian in other constructions, including the habeo construction. New nouns are recruited in every century, mostly first by the habeo construction, and later on by the mihi est construction, as follows: dor ‘longing’ (the 16th century: habeo, end of the 17th century: mihi est), teamă ‘fear’ (end of the 16th century: habeo, the 19th century: mihi est), and jenă ‘embarrassment’ (the 19th century: habeo, the 21st century: mihi est), among others. Some of the newly recruited nouns are loans from modern Romance or surrounding Slavic languages, and others are derivations of already existing, inherited nouns.

Remarkably, starting with 20th-century Romanian, the mihi est construction freely allows the occurrence of synonymous nouns in it (cf. Ilioaia 2021). For instance, the Turkish zor ‘hastiness/rush’, a synonym of the derived grabă ‘hastiness/rush’, occurs in the dative experiencer construction starting with the 20th century, and is still in use today, as opposed to grabă, which was recruited much earlier, and disappeared from it in the 21st century.

Furthermore, more recently, the mihi est construction, as opposed to the habeo construction, has shown a tendency to admit nouns from different semantic fields, such as event (plecare ‘departure’), time (atât de primăvară ‘so springtime’), and communication (cuvânt nerostit ‘untold word’), among others, as pointed out by Ilioaia (2020). These new nouns are used metaphorically in the mihi est construction and are being ‘forced’ into the construction’s meaning, that of a physiological or a psychological state, as shown in (12) and (13). This process is called coercion and is defined in the literature as the reinterpretation of a lexical item triggered by the conflict between its meaning and that of the construction in which it is used.

(12)	Mi-	e	atât	de	primăvară	încât
	me.dat=	is	so	of	springtime	that
	şoptesc	întunericului	să	[…]
	whisper.1sg	dark.the.dat	subj	[…]
	‘I feel so springtime that I whisper the darkness to (…)’
	(facebook.com/permalink.php, accessed on 5 May 2018)
(13)	mi-	e	cuvânt	nerostit…,
	me.dat=	is	word	untold
	‘I feel in a way that I cannot express (…)’
	(alexsmallthings-desprenimic.blogspot.com, accessed on 5 May 2018)

Recall that, as Lauwers and Willems (2011, p. 1220) highlighted, in Construction Grammar, coercion constitutes a major argument in favor of the existence of constructions as independent form/meaning pairings. The capability of a construction to change the meaning of a lexical item that occurs in it has been used as an empirical test to argue in favor of the construction status of structures carrying a particular meaning on their own, irrespective of the lexical items that instantiate them.

4.2. Results

4.2.1. Distribution of the Inventory of the State Nouns under Investigation

Romanian, like most other languages, has several possibilities to express psychological or physiological states by means of a structure of the type [experiencer V noun_state]. As mentioned in the previous section, besides the habeo and the mihi est constructions, there are other experiencer constructions that occur with the state nouns under scrutiny (14)–(16). Among all of the possible competing constructions, there are some that select for a dative (14), accusative (15), or nominative (16) experiencer.

(14)	Mi	s-	a	făcut	dor
	me.dat	se=	has	made	longing
	de	părinți
	of	parents
	‘I began to miss my parents.’
	(1882, Petre Ispirescu, Basme)
(15)	M-	a	cuprins	frica
	me.acc=	has	sized	fear.nom=acc
	‘I became afraid.’
(16)	Băiatul	prinse	frică	de	cățel
	boy.nom	caught	fear	of	puppy
	‘The boy became afraid of the puppy.’
(17)	Mariei	i-	e	frică	dar
	Maria.dat	her.dat=	is	fear	but
	nu	vrea	să	zică
	not	wants	subj	say.subj.3sg
	‘Maria is afraid but she doesn’t want to say it.’
(18)	După	acéia,	Gligorie-	vodă,	având
	after	that	Gligorie-	voivode	having
	grijă	și	mare	frică,	[…]
	worry	and	big	fear	[…]
	‘Then, since Grigorie-voivode was worried and very afraid, […]’
	(1659, Constantin Cantacuzino, Letopiseţul)

However, based on my data, the frequency of these other syntactic patterns is not as high as the frequency of the habeo construction (3,038 occurrences) or the mihi est construction (3,969 occurrences), as shown in (17) and (18). This can be seen in Figure 1 below, where Other refers to the other types of constructions and corresponds to the remaining 1,451 occurrences from the total of 8,458 relevant examples in my dataset, after excluding those containing the habeo or the mihi est construction.

Therefore, in what follows, I focus only on the two most frequent constructions, the habeo and the mihi est experiencer constructions. I examine how the set of state nouns that occur with them evolves throughout the centuries, aiming to make some connections between the changes in the inventory of nouns and the evolution and productivity of the constructions themselves.

4.2.2. Dynamics of the Set of Nouns Occurring in the habeo and mihi est Constructions

Remarkably, a first glance at the data reveals that, more recently, not all of the nouns that occur with the mihi est experiencer construction also occur with the habeo experiencer construction, a fact that seems to argue against the hypothesis that the state nouns occurring in the mihi est construction first occurred in the habeo construction in Romanian. In order to understand this behavior, it is important to highlight that, starting with the 19th century, the mihi est construction expands its semantic domain. Basically, as pointed out by Ilioaia (2020, 2021), the mihi est construction becomes open to synonymous nouns and to other nouns not only from the semantic field of psychological and physiological states, but also from different semantic fields, such as time (noapte ‘night’, toamnă ‘autumn’), nature (foc ‘fire’), weather (ger ‘freezing’, furtună ‘storm’), matter (vomă ‘vomit’), events (vis ‘dream’), and act/object (joc de copii ‘game of children’), among others. When part of the mihi est construction, these new nouns, most of which are used metaphorically, express a physiological or a psychological state, as shown in the examples in (19) and (20) below and in (22) further below.

(19)	Îmi	este	durere	şi	-mi
	me.dat=	is	pain	and	=me.dat
	este	furtună
	is	storm
	‘I feel pain and I feel tourmented (in my soul)’
	(reteaualiterara.ning.com/m/blogpost?id=1971741, accessed on 5 May 2018)
(20)	Mi-	e	joc	de	copii
	me.dat=	is	game	of	children
	‘I feel playful’
	(alexsmallthings-desprenimic.blogspot.com, accessed on 5 May 2018)

Not surprisingly, the nouns that recently became compatible with the mihi est construction, as well as the newly entered state nouns, are all nouns with a very low frequency, mostly up to three or four occurrences in the construction, i.e., hapax legomena (nouns that occur only once in my dataset in the construction under scrutiny), dis legomena (nouns occurring twice), or tris legomena (nouns occurring three times). Importantly, when combined with the habeo construction, these new nouns do not convey a psychological or a physiological meaning. Nevertheless, due to their common character, these nouns inevitably occur with habeo in other types of constructions, mostly conveying an abstract possession meaning or a totally different meaning not relevant here. Take, for instance, the noun toamnă ‘autumn’, a very common noun that does not belong to the semantic field of psychological or physiological states. This noun conveys an abstract possession meaning when combined with a habeo construction (21) but, when used metaphorically in the mihi est construction, as shown in (22), it expresses a psychological state.

(21)	Anul	trecut	am	avut	o
	year.the	past	have.1sg	had	an
	toamnă	ploioasă
	autumn	rainy
	‘Last year I had/ we had a rainy autumn.’
(22)	Mi-	e	toamnă,	iubito!
	me.dat=	is	autumn.nom	beloved.voc
	‘I feel melancholic, my love!’
	(intelepciune.ro, accessed on 5 May 2018)

In light of the ability of the mihi est construction to coerce nouns coming from other semantic fields into its own meaning, i.e., that of a psychological or a physiological state, the data presented in this section provide valuable evidence in favor of the construction_new status of the mihi est structure, hence, in favor of an evolution in terms of constructionalization.

4.2.3. Experiencer habeo vs. Experiencer mihi est

In what follows, I first give an overview of the competition between the two experiencer constructions, based on the entire dataset. Subsequently, I describe a few diachronic case studies in order to investigate per noun if a shift takes place from the habeo to the mihi est construction.

Figure 2 shows the competition between the two constructions when used with the state nouns in my dataset, for all periods combined. Note that, for readability purposes, the graph plots only the nouns with a relative frequency of above 0.2%. The frequencies shown in this graph have been obtained by dividing the absolute frequency of each noun by the total number of occurrences in the dataset, i.e., 7,007 occurrences. From this graph, it can be observed that the nouns occurring less frequently in the mihi est construction still have a higher frequency in the habeo construction, whereas the more frequently a noun is used in the mihi est construction, the less frequently it occurs in the habeo construction (cf. also Ilioaia 2021, p. 145).

As for the evolution of this rivalry,

Table 1. Competition of habeo vs. mihi est—diachronic perspective (cf. Table 6.4, Ilioaia 2021, p. 145).

	habeo	%	mihi est	%
16th–18th	232	3.3%	159	2.3%
19th	195	2.8%	335	4.8%
20th	301	4.3%	691	9.9%
21st	2310	33.0%	2784	39.7%
Total	3038	43.4%	3969	56.6%

offers an overview of the competition between the two constructions throughout the centuries when they occur with nouns denoting a state. Since the sets of examples for each historical period are not of equal size, the absolute figures are followed by relative figures in the table. The relative figures were calculated by dividing the absolute figures by the total number of examples in the dataset (7,007). These data are visualized in the graph in Figure 3, which illustrates that it is only in the first historical period that the habeo construction was dominant. Starting with the 19th century, the occurrence of the abovementioned state nouns with the mihi est construction prevails, and this supports the claim that these nouns were first employed in the habeo construction and then passed on to the mihi est construction, judged as more appropriate to express psychological and physiological states.

On first sight, the frequency of the nouns occurring in my dataset with the habeo construction is surprisingly high for the 21st century. However, this may be explained by the high number of new nouns entering the mihi est construction, most of which belong to the common vocabulary of the language and, hence, are very frequent. For this period, Ilioaia (2021, pp. 233–34) shows that 55% of all of the nouns occurring with mihi est have a very low frequency in this construction. They are either dis legomena (5%) or hapax legomena (50%), mostly belonging to other semantic fields, such as events, weather, time, or elements of nature, as illustrated in the graph in Figure 4.

These new nouns already occur with habeo much more often in my dataset, however, the habeo configurations into which they enter exclude a psychological or a physiological meaning. This, in fact, supports the hypothesis stated in the introduction of this paper, that the nouns under investigation first occurred in the habeo construction and shifted over time to the mihi est construction, which has become more specialized in expressing psychological and physiological states.

This evolution can also be observed diachronically, as evident from Table 1 above (cf. Ilioaia 2021), where the nouns denoting a state were more frequently found in the habeo experiencer construction in the older periods of Romanian than they were in the present-day language. Indeed, Vangaever and Ilioaia (2021) show that, in the older periods of Romanian, habeo occurred with all kinds of nouns, including concrete, abstract, and even with nouns denoting psychological or physiological states. This situation changes over time, and habeo becomes typical for expressing predicative possession1 in Romanian, while its inventory of nouns expressing a psychological or a physiological state decreases considerably, being taken over by the more specialized mihi est construction.

The evidence by Vangaever and Ilioaia (2021) reinforces the findings of an earlier case study of two of the most frequent nouns entering the mihi est construction in Romanian, frică and teamă, both meaning ‘fear’ (Van Peteghem and Ilioaia 2017). This study shows that frică was present in both constructions in the 16th century, but was more frequent in the habeo construction, whereas teamă occurred in the habeo construction at the end of the 16th century, when it was attested in the language for the first time and shifted to the mihi est construction later on, in the 19th century.

4.2.4. Case Studies

In what follows, I focus on a limited number of nouns and investigate whether these nouns occurred first in the habeo construction and then in the mihi est construction, and which of the two rival constructions is preferred throughout the centuries. The list of nouns under investigation constitutes a subset of a broader set studied by Ilioaia (2021) and contains the following nouns: frică ‘fear’, teamă ‘fear’, dor ‘longing’, scârbă ‘disgust’, poftă ‘craving’, milă ‘pity’, grijă ‘worry, care’, spaimă ‘fear’, grabă ‘rush, hurry’, and nevoie ‘need’.

Among these nouns, certain items seem to change completely in terms of their dominant construction throughout the centuries, while others continue to occur in both constructions. When the dominant construction changes, it is mostly from the habeo to the mihi est construction, and only very isolated cases show the opposite direction of change.

A closer look at the first set of nouns, namely those that show an increasing propensity toward the mihi est construction after already being attested in the habeo construction, reveals fascinating details.

For instance, two of the most frequent nouns in the mihi est construction, frică and teamă, both meaning ‘fear’, start by occurring in the habeo construction, before shifting to the mihi est construction (cf. also Van Peteghem and Ilioaia 2017). Importantly, these nouns entered into the language at different moments and from different sources, as follows: frică (<Gr. φρίϰη) was first attested in the 15th century, but might be older, whereas teamă, derived from teme ‘to fear’ (<Lat. timere), is first attested in the late 16th century.

As illustrated in Figure 5, frică, just like scârbă ‘disgust’ (<Sl. skrŭbĩ), occurred from the earliest texts in both constructions, but preferred the habeo construction during the first historical period (16th–18th), whereas teamă and, similarly, dor ‘longing’ (<Lat. dolus), appear exclusively in the habeo construction during this period. The situation changes in the 19th century for all of these nouns, when the mihi est construction becomes by far the more dominant construction with frică and scârbă, while teamă and dor start being increasingly selected by the mihi est construction. This first set of nouns may be considered the most advanced in their evolution in the mihi est construction, as their use represents the most natural way of expressing these states in Romanian.

Note that, for the visual representation of the competition between the two constructions in the graphs given in Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9, I opted for a bar chart, where each bar represents the total occurrences of that specific noun in the two constructions, for the investigated historical period. For instance, in the first graph given in Figure 5, above the first bar shows that, in the period between the 16th and 18th centuries, the noun frică ‘fear’ was more frequent in the habeo construction (blue area) than in the mihi est construction (orange area). The subsequent bars in the same graph show that the situation changes in the following centuries and that the noun becomes more frequent in the mihi est construction (orange area) than in the habeo construction (blue area).

Concerning the last noun in this group, grijă ‘worry, care’ (<Bg. griža), it has a particular evolution. After having the habeo experiencer construction as the dominant construction from the 16th to 18th centuries and as the only selecting construction in the 19th century, it totally disappears from both experiencer constructions in the 20th century, and is revived in the 21st century, where it occurs only in the mihi est experiencer construction, as illustrated in (23). My dataset for the 21st century contains 15 such examples. Although some native speakers may find these occurrences rather ill-formed, others do not think so, since the noun is employed in the mihi est construction even on official (local) news platforms. It has to be noted that, just as it is the case with other nouns in my dataset, grijă ‘worry, care’ is a very common noun in Romanian and does not completely disappear from the language during the 20th century, but only from the experiencer construction under scrutiny. Providing an explanation for this peculiar evolution is not an easy task, since its revival in the mihi est construction in the 21st century could be a recent borrowing or a modern calque of the construction as a whole. However, one should not exclude the possibility that its absence during the 20th century is a result of the scarcity of representative data for this specific historical period. In Figure 6, this evolution is visually represented.

(23)	Dar	el	a	zis	aceasta	nu	pentru că
	but	he	have.3sg	said	this	not	because
	îi	era	grijă	de	săraci,	ci
	him.dat	was	worry	of	poor.pl	but
	‘But he said this not because he was worried about the poor ones, but…’
	(gorjeanul.ro, accessed on 12 December 2023)

In what follows, a second set of nouns is investigated, containing, among others, milă ‘pity’, poftă ‘craving’, and spaimă ‘fear’. What they have in common is that their frequency in the habeo construction slightly increases or remains stable throughout the centuries, contrary to what one may expect, given the observed general tendency.

The noun milă ‘pity’ (Sl. milŭ) was attested very early in the mihi est construction (end of the 15th century). Occurrences of milă in the habeo construction are not found in my dataset before the beginning of the 17th century (cf. Figure 7). The path followed by this noun contradicts my hypothesis that the state nouns occur first in the habeo construction before being recruited by the mihi est construction. A closer look at the distribution of its occurrences reveals that milă is mostly found in translated texts, as compared to the original texts between the 15th and 16th centuries (11 occurrences in translated texts vs. 6 occurrences in original texts). This could suggest either that this noun was used in a cognate mihi est construction in the source language and has been introduced in Romanian through a loan translation of the structure, or that its absence in the habeo construction represents a consequence of the scarce data for this period of Romanian. However, this does not change the fact that this noun is first attested in the mihi est construction, and only a century later in the habeo construction.

As for poftă ‘craving’, this noun, derived from the verb pofti (<Sl. pohotĕti), shows a particular evolution, since it has a very slow increase in frequency in the mihi est construction, while habeo continues to be its dominant construction until the present-day language (cf. Figure 7).

The question arises as to whether other nouns may also show this tendency, having a low or decreasing frequency in the mihi est construction, while habeo becomes dominant. A case in point is spaimă ‘fear’, which can still be found in the mihi est construction in the 21st century, but has a considerably lower frequency, which decreases to half (see Figure 8).

The remaining two case studies refer to grabă ‘hurry’ and nevoie ‘need’, nouns that disappear completely from the mihi est construction in the 21st century and are today only found with the habeo construction. Their evolution looks similar to the evolution of spaimă ‘fear’, with the only difference being that spaimă continues to have a reasonable frequency in the present-day language with the mihi est construction. It has to be noted though, that these three nouns have never really been deep-rooted in the mihi est construction, given that their highest frequency in this construction is six occurrences for spaimă in the 19th century and only one occurrence for each of the other two nouns for the same historical period (cf. Figure 9 for grabă and nevoie).

What might the underlying factors be behind the fact that nouns such as grabă, nevoie, or even spaimă, discussed above, take such a turn and do not follow the general trend, that of changing from the habeo to the mihi est construction? A possible explanation is the competition with other synonymous well-established constructions (cf spaimă vs. frică ‘fear’) or with newly borrowed elements (cf. grabă vs. zor ‘rush, hurry’). Yet, in the case of nevoie, which occurs in my corpus only once in the mihi est experiencer construction, in a translation from the 16th century, the explanation could be either that it was an error made by the translator or a literal translation of a structure from the source text, which did not reoccur in the construction in the following centuries. However, one cannot ignore that, while most of the state nouns under scrutiny predominantly occur as bare nouns and without complementation (cf. ?am/mi-e frică (de tine) ‘I am afraid (of you)’), nevoie seems to require the presence of a complement in order to be well-formed, even in the habeo construction (cf. am/?mi-e nevoie *(de tine) ‘I need (you)’). This last point may uncover another reason why nevoie was not perceived by the speakers as being well-formed in the mihi est construction but was considered as more compatible with the habeo construction.

Given the tendencies observed, the question arises as to how many of the most frequent nouns in the 21st century have the habeo construction as the dominant construction. To obtain such a ranking, I calculate the relative frequencies in present-day Romanian of all of the nouns occurring in the two constructions taken together and sort them by their total (cf. Figure 10). The relative figures were calculated by dividing the absolute frequencies by the total number of examples in the dataset (7007 occurrences).

Seven out of the ten nouns visualized in the graph in Figure 10 have mihi est as the dominant construction to express states, whereas three of them prefer the habeo construction. One of these three nouns is nevoie ‘need’, which, as already mentioned, has been found in my dataset only once in the mihi est construction, in a translated text in old Romanian, and, hence, may be considered as an accidental occurrence, since it did not develop further in this construction (cf. also Ilioaia 2021). The other two nouns, chef ‘mood’ and oroare ‘horror’, are both newly recruited by the mihi est construction in the 21st century. I consider that it is too early to predict the evolution of these two nouns, since they only started occurring in the mihi est construction in the 21st century, more precisely, chef with three occurrences and oroare with eight occurrences. Nevertheless, their high frequency in the habeo construction corroborates my hypothesis that the state nouns under scrutiny first occur in the habeo construction before shifting to the mihi est construction, confirmed by most of the data for the older periods of Romanian.

5. Discussion

Based on the evolution of the state nouns with respect to the two competing constructions, experiencer habeo and experiencer mihi est, presented in the previous section, I suggest that, once a noun has been recruited by the mihi est construction, one of the following paths will be taken:

(i)

The state noun decreases in frequency in the habeo construction (immediately or within a century) or disappears completely from it, while it increases in frequency in the mihi est construction. This is the path followed by the following nouns: frică ‘fear’, teamă ‘fear’, dor ‘longing’, scârbă ‘disgust’, and grijă ‘worry, care’.

(ii)

The state noun keeps occurring in both constructions throughout the centuries, with the habeo construction still being the dominant one. This is the case for milă, ‘pity’, poftă ‘craving’, and spaimă ‘fear’.

(iii)

The state noun occurs for a period in both constructions but continues to increase in frequency in the habeo construction while decreasing in frequency in the mihi est construction. This evolution has been observed for grabă ‘rush, hurry’, and nevoie ‘need’.

In an attempt to provide a theoretical explanation for each of these three paths, I propose to look at the evolution of a filler (i.e., the noun denoting a state in our case) with respect to a specific slot (i.e., the opening or the position of that noun in the argument structure) as corresponding to a certain degree of entrenchment of that specific pattern in the network. The entrenchment is caused by repeated usage and often, as a result, has a high degree of schematization.

This being said, the three paths identified in the evolution of the inventory of nouns in the habeo and the mihi est constructions may be seen as different phases in the process of schematization and, hence, the constructionalization of the constructions. The nouns following the first path are among the most frequent ones and show a high degree of entrenchment in their use with mihi est, given their ‘fixation’ in the construction. The second path corresponds to a transition phase, which may result in the entrenchment of the nouns in the mihi est construction or may lead to their elimination from this construction and to their ‘fixation’ in the habeo construction. The third path corresponds to a phase in which the state noun becomes more preferred in the habeo construction than in the mihi est construction, with which it eventually becomes semantically incompatible.

On a more general note, from these case studies, I conclude that most of the state nouns under scrutiny tend to first occur in the habeo construction and then in the mihi est construction, with the only exception being the noun milă ‘pity’, which is first attested in my dataset in the mihi est construction. For some nouns, the change occurs very abruptly, whereas, for other nouns, it takes centuries. The change is, for some nouns, definitive and in favor of one of the two constructions, whereas other nouns continue to occur in both constructions.

This study also highlights that, although most of the nouns occurred first in the habeo construction, not all of them had deep roots in this construction when they were recruited by mihi est. A possible explanation for this is that these nouns had recently entered into the language. In contrast, some of the nouns entering the mihi est construction in the 21st century show a considerably high frequency in the habeo construction, since they were borrowed into the language one century earlier (oroare, in the 20th century), or even two or more centuries earlier (chef, in the 19th century). Given this, it is difficult to predict whether the new nouns entering the mihi est construction in the 21st century will be able to abandon the habeo construction, in which they seem rather deep-rooted. One thing is sure, though, their access to the experiencer mihi est construction has been granted. Note also that the nouns entering the experiencer mihi est construction in the 21st century and belonging to other semantic fields convey a totally different meaning when they occur in the habeo construction. I refer here to nouns like noapte ‘night’, furtună ‘storm’, or toamnă ‘autumn’, with some of them illustrated in the examples in (19)–(22) in Section 4.2.2, which, when occurring in the habeo construction, convey a possessive instead of an experiential meaning. This is a sign that the mihi est construction is now able to attract its own nouns, independent of their association with the experiencer habeo construction, and can coerce them into its psychological and physiological meaning.

Let us now look at these findings and interpret them from the perspective of the historical changes through which they may be characterized. One of the research questions stated at the beginning of this paper, was whether the special evolution of the two constructions can be described as a case of constructionalization or constructional change. An answer to this issue remains difficult to provide, given the difficulty in teasing apart the two concepts. Following the definition given by Noël (2007), who considers that we can talk about constructionalization when a structural pattern acquires its own meaning in such way that it adds meaning to the lexical elements occurring in it, there is no doubt that the experiencer mihi est construction represents a case of constructionalization, since it has the ability to coerce nouns from different semantic fields into its own meaning. Traugott and Trousdale (2013, p. 22), however, propose to qualify the change in a specific construction as a case of constructionalization only if the historical process results in the creation of a form−meaning pairing representing a new type node, which has been replicated across a network of language users. The difficulty one encounters when attempting to make such a decision comes, as I understand it, from the fact that constructionalization always implies constructional changes, which are said to occur before or/and after the creation of the new node in the network, even though the opposite may not always be true. So, how do we recognize when a construction has met such a degree of schematization that it can be considered as a new node in the network, if not when it has become so specialized that it coerces elements from other semantic fields into its own meaning?

With respect to the habeo construction, such a degree of schematization is very difficult to demonstrate in Romanian, given how spread the use of habeo is in several types of constructions. On the contrary, such an evolution is easier to show for the mihi est pattern. Based on the analysis of my data and reinforced by the recent tendency toward innovation of the mihi est construction, I argue that there is substantial empirical evidence in favor of an analysis of the experiencer mihi est construction as a case of constructionalization. The following arguments strengthen this statement:

(i)

The data provide evidence that the mihi est construction is, over the centuries, increasingly used with state nouns expressing psychological and physiological states with a temporary nuance, mostly recruited from the experiencer habeo construction, which shows an opposite tendency, with the state nouns decreasing in frequency. Their evolution provides additional evidence in favor of an analysis in terms of constructional differentiation (Vangaever and Ilioaia 2021, following De Smet et al. 2018, p. 198), since experiencer habeo continues to occur with abstract nouns but specializes in the contexts from which experiencer mihi est is excluded.

(ii)

The mihi est construction recruits nouns from different semantic fields, and, more importantly, coerces them into its own particular meaning, whether that of a psychological or a physiological state.

(iii)

The Latin experiencer mihi est pattern evolves in Romanian into a new, specialized construction, which becomes the most natural way of expressing psychological and physiological temporary states, whereas the experiencer habeo specializes in conveying the meaning of abstract possession.

Indeed, a clear tendency can be observed throughout the centuries, namely the attraction force exercised by the Romanian mihi est construction on state nouns, and especially on the state nouns occurring in the habeo construction, causing the habeo construction to become less preferred for expressing psychological or physiological states.

6. Conclusions

The aim of this paper was twofold. First, by connecting the development of the mihi est and habeo patterns throughout Latin with their evolution from Latin to present-day Romanian, this paper intended to provide a clearer view of the evolution of the two patterns and of the historical changes that they have passed through after being inherited by Romanian until the present-day language. Second, by having a closer look at the set of nouns that occur in the two constructions in Romanian, I aimed to understand how the changes in the inventory of the state nouns occurring in these constructions affect the evolution and the productivity of the constructions themselves.

With regard to the evolution of the two patterns in Romanian and to the historical changes that they have experienced throughout the centuries, the outcomes of the corpus studies show that both Latin patterns have been inherited by Romanian with their double nature: possessor and experiencer. The way that these constructions managed to find complementary niches of the language is fascinating. Possessor habeo continues to occur in Romanian with an unlimited inventory of concrete nouns, while possessor mihi est specializes in identificational clauses (cf. (Ea) mi-e cumnată, lit. she me.dat is sister-in-law ‘She is my sister-in-law’). As for the experiencer patterns, which are of a great importance in the present paper, experiencer mihi est specializes in expressing psychological and physiological states in Romanian, whereas experiencer habeo, which occurred during the older periods more often with abstract nouns from the semantic field of psychological and physiological states, becomes less preferred for expressing these states in the present-day language. Nevertheless, habeo continues to occur with abstract nouns pertaining to other semantic fields. Therefore, in terms of functional competition, the evolution experienced by habeo and mihi est can be described as constructional differentiation, as opposed to the constructional substitution specific for the other Romance languages.

The case studies, intended to reveal any particular dynamics in the inventory of the nouns occurring in the two experiencer constructions, show that, in general, the nouns that occur in the mihi est construction first occurred with experiencer habeo and that they all belong to the field of psychological and physiological states. More recently, however, mihi est occurs with nouns from other semantic fields, such as events, time, communication, or weather phenomena, and coerces them into its own meaning, i.e., that of a psychological or physiological state.

The specialization of the mihi est construction in expressing exclusively psychological and physiological states with a temporary nuance and its ability to coerce nouns from different semantic fields into the meaning of the construction, in corroboration with the fact that the mihi est construction becomes the most natural way of expressing this type of state in present-day Romanian, constitutes valuable evidence in favor of an analysis of this construction in terms of constructionalization.