Romanian plăcea ‘like’: an alternating Dat-Nom/Nom-Dat verb

1 Introduction

The hypothesis that there are non-nominative subjects in the Romanian language was first suggested in the literature more than two decades ago (Rivero and Geber 2003, 2004). Several scholars have argued for the existence of dative subjects in Romanian in general (see references below), while more recently, Ilioaia (2021, 2023) and Ilioaia and Van Peteghem (2021), following Cornilescu (2009), demonstrate beyond any reasonable doubt that the dative of the mihi est construction in Romanian behaves syntactically as a subject, with the nominative being a part of the predicate (Ilioaia 2023: 207, see also Van Peteghem and Ilioaia 2017). Ilioaia and Van Peteghem (2021) and Ilioaia (2020, 2021, 2023) establish this with a host of Keenan’s (1976) well-founded subject tests, whereas earlier studies on Romanian are essentially based on only two tests, i.e. the raising-to-subject test (Alboiu 2007; Cornilescu 2009; Geber 2006, 2011; Rivero 2009; Rivero and Geber 2003, 2004), on the one hand, and word order (Giurgea 2017: 307–308), on the other hand. Following Ilioaia (2021: 75, 246), Cornilescu and Tigău (2023) make similar claims for the dative of Dat-Nom verbs in Romanian, partly based on the same subject tests as established by Ilioaia (2021, 2023) and Ilioaia and Van Peteghem (2021).

Research on the subject behavior of oblique subjects in the languages of the world has been ongoing for almost five decades, starting with Andrews’ (1976) and Thráinsson’s (1979) early work on Modern Icelandic and Masica’s (1976), Kachru et al. (1976) and Klaiman’s (1980) work on the South Asian languages. Since then, a substantial body of work has been carried out on the behavioral properties of non-nominative subjects, showing beyond doubt that there are oblique subjects in several languages around the globe. These include Modern Faroese (Barnes 1986), Native American languages (Hermon 1985), Dravidian (Verma and Mohanan 1990), Dardic (Steever 1998), Japanese (Shibatani 1999), Greek (Anagnostopoulou 1999), Russian (Moore and Perlmutter 2000), Tibeto-Burman (Bickel 2004), Korean (Yoon 2004), German (Barðdal 2002, 2006, 2023: Ch. 6; Barðdal et al. 2019), Semitic (Landau 2009; Pat-El 2018), Cariban (Castro Alves 2018), Tsezic languages (Comrie et al. 2018), as well as several modern South-Asian languages (see Hock and Bashir 2016 for references).

Along with this strand of research, several scholars working on this issue have noted that at least some Indo-European languages have verbs selecting for a dative and a nominative argument which seem to allow for two word orders, i.e. dative-before-nominative and nominative-before-dative. Either word order appears to be equally neutral in the sense that one of the orders does not seem to be a topicalization of the other. Instead, the two opposite word orders are equally neutral in their own right. This phenomenon was first noted by Bernódusson (1982: 37–38) for Modern Icelandic, Barnes (1986) for Modern Faroese, Primus (1994: 67–68) for Modern German, Allen (1995: 107–109, 114–117) for Old English, Barðdal (1998) for the early Scandinavian languages, Anagnostopoulou (1999: 69) for Modern Greek, Haspelmath (2001: 69) for Italian, and Elens, Somers and Barðdal (2024) for Old Icelandic. For Romanian, Ilioaia (2021: 75, 246) and, following her, Cornilescu and Tigău (2023), argue for the existence of such alternating verbs in that language. Consider the examples in (1) below.

Dat-Nom

Nom-Dat

In these examples, the Romanian verb plăcea ‘like’ occurs with a dative experiencer and a nominative stimulus. Both word orders are equally neutral in the sense that they are not information-structurally marked (see Section 2 below). In (1a) the dative occurs in clause-initial position, while the nominative is postverbal. In contrast, in (1b) the nominative occurs in clause-initial position, while the dative is postverbal. Observe, also, that the semantic roles of the two arguments, the experiencer and the stimulus, are kept constant across the morphological cases in these two examples. This behavior suggests that the Romanian verb plăcea ‘like’ may indeed be an alternating verb of the type discussed above for several other Indo-European languages.

In (2001a) Barðdal showed for Modern Icelandic that either argument of alternating verbs, the dative or the nominative, behaves syntactically as a subject, based on seven different subject tests in that language, although not at the same time, of course. When the dative occupies the preverbal position it behaves syntactically as a subject and the nominative as an object. Conversely, when the nominative occurs preverbally, it takes on the behavior of the syntactic subject, while the dative behaves syntactically as an object. This led Barðdal (2001a) to argue that the relevant verbs instantiate two diametrically opposite argument structures, i.e. a Dat-Nom and a Nom-Dat argument structure, with each of these being an argument structure in its own right. For this reason, we use the labels Dat-Nom and Nom-Dat here to represent the two argument structure constructions, while we use the terms dative-before-nominative and nominative-before-dative to refer to linear word order in specific examples.

More recently, Barðdal et al. (2014, 2019) and Barðdal (2023: Ch. 6) show that such alternating verbs also exist in Modern German on the basis of six subject tests in that language. In a series of articles following up on this issue, Somers and colleagues carry out a frequency study of the two word orders in both Icelandic and German, documenting that the dative-before-nominative and the nominative-before-dative word orders show approximately a 50–50 distribution when the two arguments are full NPs. This stands in stark contrast to, for instance, verbs like ‘thank’ which select for a nominative subject and a dative object, with which the nominative-before-dative word order is found in more than 95 % of the cases under the same conditions (Somers 2024; Somers et al. 2024a, 2024b; Somers et al. 2025).

In brand-new research, Cluyse et al. (2025) also argue for the existence of such verbs in Latin, in particular for the verb placēre ‘like, please’, again on the basis of six different subject tests valid for Latin. Note that Romanian plăcea ‘like’ is a direct descendant of Latin placēre, which, in turn, raises the question of whether Romanian plăcea may have inherited this alternating behavior from its Latin predecessor.

This article is devoted to the issue of whether or not plăcea ‘like’ is an alternating verb in Romanian, i.e. whether either argument, the dative or the nominative, may behave syntactically as a subject, while the other argument behaves as an object. In order to investigate this, we discuss six behavioral subject properties which have already been established as subject diagnostics in Romanian (cf. Cornilescu and Tigău 2023; Ilioaia 2021, 2023; Ilioaia and Van Peteghem 2021) and we compare these behaviors with the syntactic behavior of both the dative and the nominative of plăcea. These subject properties are the following: word order, control into infinitives, control into subjunctives, raising-to-subject from finite clauses, raising-to-subject from infinitives, and raising-to-object.

The article is organized as follows: Section 2 starts with a brief introduction of the Romanian verb plăcea ‘like’ and its different argument structures. This includes a discussion of neutral word order as opposed to word order configurations which are information-structurally marked. We then continue, in Section 3, to a corpus study of the word order distribution found with plăcea across the two word order configurations, dative-before-nominative and nominative-before-dative. In Section 4, we provide a discussion of the concept of alternating Dat-Nom/Nom-Dat predicates, including the relation between subject status and event structure. Before we dive further into the issue of subjecthood in Romanian, Section 5 provides a short overview of some of the subject tests that have been used in recent literature, which we argue do not distinguish between syntactic subjects and objects and can therefore not be employed. Section 6 presents the set of properties we use as diagnostics for subject status in Romanian, focusing in particular on several of Keenan’s (1976) well-established behavioral subject tests. We apply these tests to the two arguments of plăcea in order to examine the extent to which the dative and the nominative behave syntactically as subjects, when placed in preverbal position. Section 7 summarizes the content of this article and our conclusions.

2 The Romanian verb plăcea ‘like’

The verb plăcea ‘like’ is one of the most frequently used psychological verbs in Romanian, derived from Latin placēre, which also means ‘like’ and sometimes ‘please’. This verb has preserved its original meaning and argument structure in Romanian until today. Similarly to its cognates in the other Romance languages (cf. Fr. plaire, It. piacere), the Romanian verb plăcea introduces a state event involving two participants, a dative experiencer and a nominative stimulus. The dative experiencer is either realized as a clitic only (2a), as a strong pronoun (2b), or as a full NP (2c). In the last two cases, the dative is obligatorily doubled by a clitic when it occurs in clause-initial position.

Dative as a clitic

Dative as a strong pronoun

Dative as a full NP

In examples (2a–c) above, the presence of the clitic is obligatory. Note, however, that this has not always been the case in Romanian, as is documented by Rivero (2009: 2) as early as for the first decennium of the 18th century. Although the absence of the clitic is considered archaic, one may still find instances of non-doubled datives in present-day Romanian. In our dataset, which contains 19 instances of non-doubled nominal datives (vs. 545 doubled datives), this appears to be confined to configurations where the dative is postverbal, i.e. occurring in the Nom-Dat construction, especially in examples where the dative expresses a generic meaning, as in (3) below (cf. also Geber 2011: 137–143).

Dative not doubled by a clitic

Turning to the stimulus, this argument may be realized either as a full NP or as a strong pronoun in the (unmarked) nominative case, as in the examples in (1) above, as a prepositional object (4a), a finite subordinate clause (4b) or a non-finite clause (4c). When the nominative stimulus refers to an already mentioned referent, as in (3) above, it remains unexpressed, since Romanian is a pro-drop language and does not have nominative clitics.

Dat-PP

Dat-Clause

Dat-Inf

At this juncture, it is appropriate to mention that there appears to be a slight difference in the meaning of plăcea ‘like’ depending on whether it instantiates the Dat-Nom or the Nom-Dat construction, yielding the meaning ‘like’ in Dat-Nom constructions but ‘be to sb’s liking’ in Nom-Dat constructions. This is essentially the same lexical meaning, only involving a difference in perspective, concomitant with the two argument structures. Hence, in the remainder of this article we gloss and translate plăcea as either ‘like’ or ‘be to sb’s liking’, depending on our analysis of the relevant argument structure in each example.

Besides the argument structures involving a dative experiencer discussed above, plăcea ‘like’ has developed a new argument structure in Modern Romanian, where the two core arguments are canonically marked, i.e. the Nom-Acc case frame (cf. Ilioaia 2022). An example is given in (5) below, where the experiencer cineva ‘somebody’ is encoded in the nominative case and the stimulus te ‘you’ in the accusative.

Nom-Acc

This argument structure was first attested at the end of the 18th century and remains relatively rare until the present-day language (cf. Ilioaia 2022).^[1] This particular argument structure will not be further investigated here.

Turning now to the word order variation found with plăcea ‘like’ in (1) above, both word orders appear to be equally neutral in the sense that, according to native speakers, they are not information-structurally marked. For a comparison, consider the following two examples with the Romanian mulțumi ‘thank’, a verb which also selects for a nominative and a dative argument. The example in (6a) represents neutral word order in Romanian, with the nominative subject in first position and the dative object in postverbal position.

Nom-Dat

Dative-before-nominative – contrastive focus

In contrast, the dative-before-nominative word order with mulțumi ‘thank’ in (6b) is only fully grammatical with a contrastive focus reading, as is evident from the translation into English with a cleft construction. In other words, even though the dative of mulțumi may precede the nominative, this does in no way represent neutral word order, as is the case with plăcea ‘like’ in (1) above.

To conclude, this comparison between the word order distribution of the nominative and the dative with plăcea ‘like’ and the nominative and the dative with mulțumi ‘thank’ has confirmed that an alternation in the word order between nominative and dative arguments is not a general property of all verbs selecting for such arguments. Instead, it appears to be specific for plăcea, as opposed to mulțumi. In Section 6 below we investigate whether and to which extent either argument of plăcea, the dative and the nominative, passes the subject tests in Romanian and thus whether plăcea may instantiate the two diametrically opposite argument structure constructions, Dat-Nom and Nom-Dat, a systematic alternation also documented in several of its Indo-European cousin languages. Before that, however, we present a corpus study of the word order distribution found with plăcea in present-day Romanian texts.

3 Corpus study of word order: plăcea ‘like’

Turning now to word order, there appears to be little consensus in the Romanian scholarship on which word order is neutral and which is restricted by information-structural considerations (Giurgea 2017; Nicolae 2019; Pană Dindelegan 2013, 2016; Vasilescu 2013: 537). Yet, based on a corpus of earlier periods of Romanian, Pană Dindelegan (2016; cf. also Nicolae 2019), argues that the canonical subject is mostly preverbal in main clauses, whereas in subordinate clauses more variation is found with regard to word order (cf. Pană Dindelegan 2013: 121, 2016: 112). Similar research for 20th century Romanian has confirmed this (Giurgea 2017: 292).

For the present-day language, the only study we are aware of is Ilioaia (2023: 154–158), based on a dataset of 2,458 instances of argument structure constructions involving the Nom-Acc case frame of 28 canonically case-marked verbs with a nominative experiencer subject. The dataset includes tokens where the nominative is realized either by a strong pronoun or by an NP (662 examples, 26.9 %) as well as tokens where the nominative is a pro, hence, not realized (1,796 examples, 73.1 %).

Restricting our presentation here to the instances where both arguments are expressed, Table 1 shows that, among the 662 examples with the nominative subject realized by either a strong pronoun or by an NP, 95.8 % of the tokens have the subject in preverbal position while the subject occurs postverbally in only 4.8 % of the cases. Thus, Ilioaia’s (2023) findings are in accordance with the conclusions of earlier studies, namely that nominative subjects tend to occur in preverbal position, whereas the object prefers the postverbal position.

Given this, we take SVO to be the canonical word order in Romanian. For the verb plăcea ‘like’, we predict that both the dative-before-nominative and the nominative-before-dative orders are SVO structures in Romanian, instantiating the two argument structures, Dat-Nom and Nom-Dat, respectively, instead of one of the word orders being a topicalization of the other.

In order to verify the aforementioned hypothesis, we have carried out a synchronic corpus study, providing word order statistics for the verb plăcea ‘like’, and as a control, also for the verb mulțumi ‘thank’, which is an unambiguous Nom-Dat verb. Thus, we have collected material for both verbs from the Romanian Web Corpus roTenTen16 (Jakubíček et al. 2013), which consists of 2.6 billion words. We have extracted a sample including the first 200 main clause tokens for each of the two verbs, where these occur with two arguments, a nominative and a dative, at a distance of zero-to-three words. We have only included tokens where the two arguments are represented as either full NPs or strong pronouns, with one of the arguments being preverbal and the other postverbal. Examples where either of the two arguments are pro-dropped are excluded for the simple reason that the arguments are not expressed in such examples, which in turn makes it difficult to exactly decide on their position in the clause.

Our findings for mulțumi ‘thank’ are shown in Table 2, establishing beyond doubt that mulțumi is a Nom-Dat verb in Romanian, as all 200 tokens exhibit the nominative-before-dative word order. However, as we discuss in Section 2 above, the reverse order, dative-before-nominative is not ungrammatical in Romanian, despite none such examples being among the 200 extracted tokens of mulțumi in our dataset. This may be validated with a quick Google query, which certainly yields examples of mulțumi ‘thank’ containing a topicalized strong dative pronoun in clause-initial position, of which two are shown in (7) below:

Dative-(before-nominative) – focus

Both (7a) and (7b) are clear examples of focus, as is evident by the use of the cleft in the translation of these examples into English and further corroborated by the use of the focus particle tot ‘once again’ in clause-initial position in (7a). Note that the nominative is pro-dropped in both cases, which is most likely the reason why this word order does not show up in the statistics in our corpus study, as it excludes pro-dropped arguments.

Turning to plăcea ‘like’, Table 3 shows that there is a strong bias towards the use of the dative-before-nominative word order, at the cost of the nominative-before-dative word order, in all configurations except when both arguments are pronouns. In that last configuration, the distribution between the two word orders is 87.5 % versus 12.5 %. Taken together, these numbers rather suggest that plăcea is a Dat-Nom verb and not necessarily an alternating Dat-Nom/Nom-Dat verb, contrary to our claim that both word orders are neutral, as is the case in the example pair in (1).

Nevertheless, there is still a major difference between plăcea ‘like’ and mulțumi ‘thank’ in that dative-before-nominative orders with mulțumi, as in (6b) above, are clearly topicalizations, while the four nominative-before-dative examples of plăcea from our corpus extract do not show any such behavior, although the nominatives are undoubtedly definite, yielding old information. These four examples are produced in (8) below:

Nom-Dat

According to our informal inquiries, native speakers agree that all four Nom-Dat examples of plăcea ‘like’ in (8) above represent neutral word order, as there are no specific information-structural restrictions at play here. These considerations, however, do not exclude either argument with plăcea from being focused, as this can be done with intonation in Romanian, irrespective of word order (cf. Göbbel 2003).

To conclude, while we have been unable to verify with corpus statistics that the two word order configurations with plăcea ‘like’, dative-before-nominative and nominative-before-dative, represent two different argument structures, i.e. Dat-Nom and Nom-Dat, we still maintain that the two word orders in Romanian are equally neutral with the verb plăcea. This we have shown through a comparison of unambiguous focalized dative objects in clause-initial position with the verb mulțumi ‘thank’, as opposed to the four nominative-before-dative orders of plăcea from our corpus extract, which all turn out to represent neutral word order. In fact, this calls for a meticulous investigation of how the two arguments of plăcea fare with regard to the subject tests in Romanian. Before that, however, a few words on the issue of subjecthood with alternating predicates are in order.

4 Alternating predicates and subject status

As already mentioned above, verbs exhibiting the Dat-Nom/Nom-Dat alternation have been documented in several Indo-European languages, both modern and ancient. In addition to Modern Icelandic, Modern Faroese, Modern German, Old English, Latin, and the history of the Scandinavian languages, Barðdal (2023: Ch. 3) presents data compatible with the hypothesis that alternating structures of this kind existed in Gothic, Ancient Greek, Sanskrit, Hittite and Old Russian, as well as in Modern Russian and Modern Lithuanian. We do not discuss these data any further here, but concentrate instead on the situation in present-day Romanian.

Having shown above that the alternation between the dative-before-nominative and the nominative-before-dative orders with plăcea ‘like’ are equally neutral in Romanian, the question arises as to why plăcea allows for this neutral word order alternation to begin with, while mulțumi ‘thank’, for instance, does not. We suggest that the reason is that plăcea is force-dynamically neutral in the sense of Talmy’s (1985, 1988) force-dynamics, while the same does not apply to mulțumi.

As outlined by Croft (1998, 2001, 2012), the assignment of subject and object status is a direct consequence of verbal semantics and the relevant force-dynamics involved. For instance, for a causative verb like kill, there is an initiator of the event, the killer, who exerts force on an endpoint of the event, the killed entity. Due to this inherent causative structure of kill, the initiator of the event, the one carrying it out, becomes the syntactic subject, while the endpoint of the event, the one acted upon, becomes the syntactic object. The verb thank, although not causative in the sense of kill, still has the same type of force-dynamics in that there is an initiator, the thanking person, who carries out the event, and there is an endpoint, the person being thanked. As a consequence, the thanking person becomes the syntactic subject, while the person being thanked becomes the syntactic object.

No such causal chain, as is found with kill, is a part of the event structure of experiencer verbs (cf. Barðdal 2023: Ch. 2). This is a consequence of the fact that experiencer verbs are susceptible to two different construals of the same experiencer event; either a construal in which the experiencer directs his/her attention towards a stimulus, resulting in the experiencer becoming a subject and the stimulus an object, or a construal in which a stimulus affects an experiencer, resulting in the stimulus becoming a subject and the experiencer an object. These two “opposite” construals are lexicalized through the two verbs, fear and frighten in English and other languages. With the verb fear, the experiencer is the subject and the stimulus the object, while with the verb frighten, the syntactic roles are reversed.

Shifting our attention to alternating Dat-Nom/Nom-Dat verbs, these verbs select for a dative experiencer and a nominative stimulus and they are either stative or inchoative in nature (cf. Barðdal 2001a, 2023: 81–84; Somers 2021). As such, they are indeed force-dynamically neutral, as already mentioned. That is, these verbs are susceptible to either of the two construals referred to above, the fear construal or the frighten construal, although not at the same time, of course. As a consequence, alternating predicates may instantiate two diametrically opposed argument structure constructions, the Dat-Nom construction and the Nom-Dat construction. Such an analysis predicts that the dative should pass the subject tests when it occurs preverbally, just as much as the nominative should pass the subject tests when it is in preverbal position.

The remainder of this article is devoted to testing the hypothesis that the Romanian verb plăcea ‘like’ is an alternating verb of this type, instantiating two diametrically opposed argument structure constructions, Dat-Nom and Nom-Dat, to be corroborated by the fact that the dative in Dat-Nom structures and the nominative in Nom-Dat structures pass the subject tests applicable in Romanian. Before that, however, a few words are in place on which syntactic property is, and which is not, a genuine subject test in Romanian.

5 Subject tests: non-applicability in Romanian

In this section we discuss a set of subject tests that are not applicable in Romanian, due to the nature of Romanian grammar. These are subject-verb agreement, deletion in imperatives, conjunction reduction, ellipsis in telegraphic style, and binding of reflexives. We address each of these below, to different degrees.

It is well known that canonical subjects in Romanian are encoded in the nominative case and trigger subject-verb agreement. However, these coding properties are not found with subject-like accusatives or datives in Romanian or in any other language for that matter. Rather, it has been argued in the literature on the Indo-European languages which exhibit non-nominative subjects that verb agreement is a property confined to nominative case marking, irrespective of whether this nominative is syntactically a subject or an object (cf. Barðdal 2023: 51–52; Barðdal and Eythórsson 2003a, 2003b; Eythórsson and Barðdal 2005; Sigurðsson 1990–91, 2004, inter alia). In other words, verb agreement is essentially a nominative test, not a test singling out syntactic subjects. As a consequence, non-nominative subjects either occur with a default 3rd person singular marking, as in examples (2–3) above, (cf. Corbett 1991: 204) or with 3rd person plural marking when the nominative argument is plural, as in the examples in (1) above. However, since verb agreement is confined to nominative arguments, be they syntactic subjects or objects, and is not found with oblique subjects, we do not discuss subject-verb agreement any further here.

One behavioral subject test, often discussed in the literature on subjecthood, deletion in imperatives, also turns out to be confined to nominative subjects, exactly like verb agreement, for the simple reason that non-nominative subjects are generally not agentive enough to occur in imperatives (cf. Barnes 1986: 25 on Modern Faroese, Rögnvaldsson 1996: 48 on Old Icelandic, and Barðdal 2006: 54 on Modern Icelandic and Modern German). Thus, this test will not be discussed any further here.

Finally, two additional behavioral subject tests, also repeatedly found in the discussion of subject status in the literature, are conjunction reduction and binding of reflexives. It turns out that neither of the two conclusively distinguish between subjects and objects in Romanian for reasons to be discussed in 5.1 and 5.2 below (cf. also Ilioaia 2023: 161–166, 183–184). Thereafter, in Section 6, we proceed to the five subject tests that genuinely tease apart subjects from objects in Romanian.

6 The status of plăcea ‘like’ as an alternating Dat-Nom/Nom-Dat verb

The subject tests relevant for Romanian have been established on the basis of Keenan’s (1976) behavioral properties of subjects. These include raising-to-object, raising-to-subject out of infinitives, raising-to-subject out of subjunctives, and control, i.e. the ability to be the controlled argument in infinitives and subjunctives. A similar set of diagnostics was convincingly applied by Ilioaia and Van Peteghem (2021) and Ilioaia (2021, 2023: Ch. 8) to argue that subject-like accusatives and datives in non-canonical subject constructions pattern with canonical nominative subjects in this language (cf. also Cornilescu and Tigău 2023).

Following the suggestions of Ilioaia (2021: 75, 246), we investigate below to which extent the two arguments of the verb plăcea ‘like’, i.e. the dative and the nominative, pass the behavioral subject tests when occurring in preverbal position. This we do in order to establish whether or not the Romanian verb plăcea is an alternating verb, instantiating two diametrically opposed argument structures, the Dat-Nom and the Nom-Dat constructions.

We start with two different control constructions in Romanian (6.1), i.e. control into infinitive complement clauses and control into finite clauses. We then proceed to raising-to-subject out of infinitives and out of subjunctives in 6.2 and raising-to-object in 6.3.

7 Summary and conclusions

The hypothesis launched and argued for in this article is that the Romanian verb plăcea ‘like’ is an alternating Dat-Nom/Nom-Dat predicate instantiating two diametrically opposed argument structures, i.e. the Dat-Nom and the Dat-Nom argument structure constructions, and that this behavior is inherited from its Latin ancestor, placēre meaning ‘like, please’. This hypothesis may be directly derived from the event structure of this type of experiencer predicates which are stative or inchoative in nature; in other words, they are force-dynamically neutral. As a consequence, two types of event construals are available to these verbs, involving i) the experiencer directing his/her attention to the stimulus, and ii) the stimulus affecting the experiencer. The former type of construal yields a Dat-Nom argument structure, while the latter type yields a Nom-Dat argument structure construction for verbs like plăcea.

This hypothesis, that plăcea ‘like’ is an alternating Dat-Nom-/Nom-Dat predicate, makes certain predictions about the syntactic behavior of the two arguments, namely that either argument, the dative or the nominative, may behave syntactically as a subject and the other argument as an object, although not at the same time, of course. When the verb plăcea instantiates the Dat-Nom argument structure, it is the dative that takes on subject behavior, while the nominative behaves syntactically as an object. In contrast, when plăcea instantiates the Nom-Dat argument structure construction, it is the nominative that behaves syntactically as a subject and the dative as an object.

In order to verify these predictions, we have carried out a word order study for plăcea, ‘like’, also including mulțumi ‘thank’ as a control. Thus, we have extracted 200 main clause examples of plăcea and mulțumi from the Romanian Web Corpus roTenTen16, examples where both arguments, the dative and the nominative, are expressed in order to determine their internal word order and their relative frequencies. Our findings, indeed, corroborate a subject analysis of the dative in Dat-Nom constructions, while at the same time, the few instances of nominative-before-dative orders are clearly not topicalizations, but represent neutral word order. This stands in stark contrast to the word order distribution found with mulțumi ‘thank’, with which the dative-before-nominative orders are clearly topicalizations, thus not representing neutral word order. These facts call for a further scrutiny of the available subject tests in Romanian, recently suggested in the literature (cf. Cornilescu and Tigău 2023; Ilioaia 2021, 2023; Ilioaia and Van Peteghem 2021).

Furthermore, we have found that two of the suggested subject tests, conjunction reduction and binding of reflexives either do not distinguish between subjects and objects or are problematic in other respects and are thus not applicable. First, not only may subjects bind reflexives, but also indirect and direct objects. For this reason, we have not relied on binding as a subject test here. Second, conjunction reduction is also problematic, on the one hand, due to the obligatoriness of the clitics, and, on the other, due to the pro-drop property of Romanian, yielding examples allegedly representing conjunction reduction as equally well explainable as pro-drop. The pro-drop property of Romanian also rules out the applicability of the test involving ellipsis in telegraphic style.

In contrast, we have argued that the subject tests safely applicable in Romanian are: control into infinitives, control into subjunctives clauses, control into adjunct gerund clauses, raising-to-subject and raising-to-object. We have shown here that uncontroversial syntactic subjects, pattern with ordinary nominative subjects, in that they pass these tests while the object stays in its position, exactly like objects do in finite clauses. Applying these five subject tests on the two arguments of plăcea ‘like’ also reveals that our hypothesis that plăcea is an alternating Dat-Nom/Nom-Dat predicate in Romanian is borne out, as either argument, the dative or the nominative, passes these five subject tests. Moreover, when the dative behaves as a syntactic subject, the nominative takes on object behavior, and vice versa, when the nominative behaves as a subject, the dative takes on the behavior of object.