Quirky subjects in nonactive sentences in Albanian

In this paper, I explore the syntax of two quirky Albanian subject constructions, both containing a dative experiencer that surfaces in subject position and behaves like a subject even if it does not induce agreement with the verb, along with an object bearing the nominative case, triggering agreement on the verb. The ﬁ rst type of quirky subject construction is characterised by the restriction of verb agreement to the third person, while the second type does not show person restrictions. The partial agreement data discussed here can be accounted for by resorting to the Person-Case Constraint. A split-feature checking analysis is conducted, whereby the person feature is checked separately from the number feature in contexts where a dative and a nominative DP are associated with two di ﬀ erent features of a unique single probe (Tense). (cid:1) 2024 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http:// creativecommons.org/licenses/by/4.0/).


INTRODUCTION
In this paper, I investigate two little-studied Albanian quirky subject constructions involving an experiencer and a theme.Both constructions contain quirky dative case-marked elements that occupy the canonical clausal subject position, even if they do not bear the nominative case and nominative objects that appear postverbally and agree with the verb.The first type of construction is represented by structures in which the nominative object and the verb have a person restriction: they can be only third-person singular or plural.A similar restriction was observed in Icelandic quirky subject constructions (Sigursson, 1992(Sigursson, , 1996(Sigursson, , 2000;;Taraldsen, 1995) and is reminiscent of Bonet's (1994) Person-Case Constraint, according to which, in the presence of an accusative and a dative, the accusative must be the third person.The second type of Albanian quirky subject construction has no person restrictions.Albanian data show that verbal agreement is uniquely associated not with the logical subject but with the nominal displaying nominative case, regardless of its grammatical function: a direct object can control agreement if it is nominative.
Albanian quirky subject constructions are allowed only in structures containing verbs in a nonactive form, the same morphological verb form that characterises passive sentences.However, in this work I show that the syntax of quirky subject constructions does not coincide with that of the passive.In the passive, the internal argument of the corresponding active sentence is raised to the subject position (SpecTP) because it lacks a case value in its base position (passives are h-incomplete verbs unable to check the accusative case of the object, which thus takes the nominative form T).
In quirky subject constructions, on the other hand, it is the dative DP that moves to the SpecTP to check the EPP feature of T, whereas the object appears in its base position following the dative DP.
Concerning the obligatory third-person agreement on the verb, which characterises one of the two types of Albanian quirky subject constructions, I assume a split-feature checking analysis whereby the person feature is checked separately from the number feature in contexts where a dative and a nominative are associated with two different features of a unique single probe (T).
The phenomena associated with quirky subject constructions relate to general theoretical topics, such as case assignment, agreement, defective intervention, and person restrictions.
The paper is organised as follows.In Section 2, I provide a simple overview of Albanian active and nonactive sentences.In Section 3, I present the central facts of Albanian quirky subject constructions and provide examples of both types, with and without person restrictions.I present a series of tests that demonstrate that Albanian quirky dative DPs have syntactic properties in common with nominative subjects.In Section 4, I consider the differences between passives and quirky subject constructions and the case-checking mechanism of quirky subject constructions.In Section 5, I address the question of how person and number features are checked by probe T.

SENTENCE STRUCTURE IN ALBANIAN
Albanian is a nominative-accusative language with five morphological cases: nominative, accusative, genitive, dative and ablative.The nominative case is normally assigned to the subject, the accusative case is assigned to the direct object, the dative case is assigned to the indirect object, the genitive case is used to express possession, and the ablative case is used (following some prepositions) to express motion, location, and time.The unmarked word order is SVO, both in main (1a) and embedded (1b) clauses: (1) a.
Meri In Albanian, the nominative case is also assigned to the subject of a passive sentence.In this case, the agent is expressed through a byphrase (3b): (i) roman 'novel' (ii) romani 'the novel' The -n ending in romanin corresponds to the accusative case.
Beni *Beni foli mua The syntax of clitics is not the focus of this paper and is thus not further developed here.My proposal about cliticization is closely related to that of Matushanski (2006), who assumes that cliticization results from the interaction of two operations: a) internal merge, which is a syntactic operation that moves the clitic from its argument position and adjoins it to the edge of T after the main verb has moved to T; and b) MÀmerger, which is a morphological operation that applies to heads and creates complex nodes (Clitic + Verb) that are syntactically atomic (Turano, 2012(Turano, , 2017)).According to this view, clitics are DPs moved from argument positions and adjoined to the main verb after it has moved to and MÀmerged with T°.Schematically, the derivation is as follows: (6) First, V raises and merges with T. Afterwards, the clitic raises and MÀmerges with the derived V + T head.The MÀmerger applies to adjacent heads and forms a new complex head, Cl + V, which acts as a single constituent.
In transitive double-object constructions hosting two internal arguments, an accusative DP can precede a dative DP (7a) and vice versa (7b 3 The fact that only the direct object can be promoted to the subject position in passives is typologically very widespread.
In summary, the nominative case is the default case for subjects, the accusative case is the default case for direct objects, and the dative case is the default case for indirect objects.The verb agreement is with nominative subjects.
A crucial property of the sentences under discussion is the nonactive form of the verb.Before proceeding with the data, I briefly present the Albanian nonactive morphology.
The Albanian language has two distinct voice paradigms: active and nonactive.The nonactive voice is realised in three distinct ways that have a well-established distribution: 1) By means of affixes attached to the verb, in the present or in the imperfect of the indicative, subjunctive and conditional moods.This strategy is exemplified in (12a) and (12b), which illustrate the indicative present and imperfect tenses, respectively.The nonactive forms can be compared to the active forms in (13a-b): ( In (17d), the internal argument behaves as a subject, while the external argument is realised as an agentive byphrase.
In the next section, I present a particular type of construction characterised by the presence of the nonactive verbal voice, where external and internal arguments receive a morphological case that is different from the case they receive in regular transitive sentences such as those illustrated in (1).

DATIVE SUBJECTS AND NOMINATIVE OBJECTS
In addition to regular passives, Albanian has another type of construction in which the morphological case does not match grammatical functions.In these constructions, which have a nonactive verb, the logical subject is an experiencer and appears in the dative case, whereas the object takes the nominative case.This type of construction includes both transitive verbs, such as bëj 'do', ha 'eat', pi 'drink'/'smoke ', punoj 'work', 7  As these examples show, the translation of the sentences corresponds to 'X wants to. ..'/'X feels like. ...' The same verbs can be found in structures that do not have the 'want' meaning.Thus, the structures in ( 19) alternate with the regular transitive structures in (21), with an active verb, a nominative subject and an accusative object, whereas the structures in ( 20) alternate with the intransitive structures in ( 22). ( 21 Thus, the Albanian dative constructions of the type illustrated in ( 19) and ( 20) display a special behaviour that differentiates them from regular transitive/intransitive sentences, such as those illustrated in ( 21) and ( 22), where a) the verb has an active form; b) the subject has a nominative case and precedes the verb; and the object has an accusative case.In ( 19) and ( 20), the logical subject appears in the dative case, the object has a nominative case, and the verb has a nonactive form.
Structures characterised by a dative subject and a nominative object, also known as involuntary state constructions, are found in Spanish (Masullo, 1993;Rivero, 2004Rivero, , 2009)), Bulgarian and Slovenian (Rivero, 2009). 8lbanian has two classes of constructions that take dative subjects and nominative objects.The first class includes verbs such as lakmoj 'to long for'/'covet', teket 'fancy', and vesk 'seduce'/'tempt' and shows a person restriction reminiscent of Icelandic quirky subject constructions.These structures, exemplified in ( 23)-( 24), contain a dative logical subject and a nominative logical object that is in the third person, singular or plural; firstor second-person nominatives are unacceptable.The verb can be only a third person, and it triggers agreement on the logical object.The dative subject does not trigger agreement on the verb.( 23 In summary, Albanian presents two patterns with dative subject constructions: one type displays person restrictions, and the other type has no person restrictions.Both types are characterised by a nonactive verbal morphology and by the logical subject never determining subject-predicate agreement.The verb always triggers agreement on the logical object that appears in the nominative.Furthermore, in dative subject constructions, nominals bearing the nominative and dative cases appear in positions that differ from those found in unmarked transitive sentences, as the nominative case is assigned to the DP in the postverbal position, whereas the dative DP appears in the preverbal position, which is usually occupied by the nominative subject (cf.(1a)).However, a nominative object can occupy the sentence-initial position if it is topicalized, as in ( 27 Albanian constructions containing dative subjects and nominative objects show the same properties as the Icelandic quirky subject constructions analysed by Zaenen et al. (1985), Taraldsen (1995), and Sigursson (1996Sigursson ( , 2000)).Icelandic has passive/unaccusative sentences in which the subject is dative and verbal agreement is with a nominative object that must be in the third person and cannot be in the first or second person.Sentences containing first or second nominative objects are ill formed (examples from Sigursson (2000: 87) Spanish also has quirky constructions displaying person restrictions identical to those of Icelandic (Masullo, 1993;Rivero, 2004).For example, the verb antojar (se) 'fancy' takes a dative subject and a nominative object that triggers verbal agreement restricted to the third person.The first or second person is ungrammatical (examples from Rivero, 2004: 496) The Albanian sentences in ( 23)-( 24), with person restrictions on the nominative object, seem to be similar to those of Icelandic ( 29) and Spanish constructions in (30), which have identical person restrictions.
Albanian also has dative constructions without person restrictions, such as those illustrated in (25), which resemble Spanish sentences containing the verb gustar 'like' (examples from Rivero, 2004: 495) For Icelandic, the tests in (32) have been used to show that quirky subjects behave like nominative subjects with respect to a series of syntactic phenomena (Andrews, 1982;Zaenen et al., 1985;Sigursson, 1992Sigursson, , 2000;;Boeckx, 2000) With respect to the contexts shown in (32), Icelandic oblique subjects are similar to standard nominative subjects.
Albanian passes five of the tests listed in (32), showing that dative subjects have properties typical of nominative subjects.For example, with respect to reflexivization (32a), Albanian dative subjects can be binders for anaphors; this is shown in (39), where the datives mua and Benit are the antecedents of the reflexive nominative vetja 9 : 9 An anonymous reviewer asks whether the nominative form of the reflexive pronoun vetja is only used in these contexts.Actually, vetja can also appear in other structures and with other morphological cases.In (a) we can see the nominative form; in (b), the accusative and in (c) the dative: (i) a.
Vetja The test in (32c), which shows the behaviour of the subject in ECM infinitives, cannot be applied to Albanian since this language has no infinitive ECM constructions.
Albanian dative subjects also satisfy criterion (32d): in raising structures, it is the dative DP that raises with the raising verb. 11 'Ben looks at the photo and remembers you on the beach.' In both sentences, Beni is the antecedent of the covert dative subject of the coordinate construction.
In Spanish, Masullo (1993) also showed that preverbal datives behave like subjects in several respects.In particular, the properties that dative subjects share with canonical subjects are listed in ( 44 The following examples, taken from Masullo (1993: 306-308), illustrate each of these properties.With respect to word order (44a), quirky subjects, like canonical subjects, can occur in embedded clauses (45a), whereas in questions, they occur in the postverbal position (45b): (i) At the same time, as Masullo (1993: 309-311) shows, there is evidence that Spanish is different from Icelandic.Spanish dative subjects cannot be represented as PRO; they cannot be deleted under identity with a nominative subject; and they are not able to bind anaphors.
Albanian also passes the tests that Masullo used to show that Spanish dative subjects and canonical subjects have similar functions.Thus, Albanian dative DPs can occur in embedded clauses (51)  On the basis of all of these tests, I conclude that the dative DP in Albanian constructions is a subject.I therefore adopt the term quirky subject to define this DP, and I call Albanian constructions containing dative subjects and nominative objects quirky subject constructions.
For Albanian, Kallulli (2006) describes structures containing a dative subject in her work on nonactive verbal morphology.She refers to these constructions as structures that have an involuntary state reading.In this analysis, structures with a dative subject are derived from dyadic agentive activity predications.Agentive activity predicates contain two features: [+intent] and [+activity].When the feature [+intent] in v is suppressed, no agent argument can be realised in SpecvP.However, v has a feature [+activity] that needs to be saturated, and this can be done only by the dative argument.When the nonactive morphology suppresses the feature [+intent] in v, the dative moves from SpecVP to SpecvP to licence the [+activity] feature.This analysis can explain why the dative element occupies the canonical subject position but fails to account for how the dative case is assigned, why the dative element never agrees with the verb, why the logical object is marked with a nominative case, and why some constructions display person restrictions.In brief, these characteristics of Albanian dative subject constructions have never been studied before.
As shown in Section 3, dative quirky subjects in Albanian appear in sentences containing a nonactive verb.In particular, these sentences show the characteristics of passives: they have a nonactive verbal form and a logical object marked with a nominative case.However, the similarity with passives is only partial because passive morphology in Albanian also occurs with the following syntactic phenomena: a) the object raises to the subject position, where it takes a nominative case and controls person agreement on the verb; b) the logical subject surfaces as an optional agentive byphrase.Therefore, passives have preverbal nominative objects and postverbal byphrase subjects.In quirky subject constructions, instead, the nominative object appears postverbally, whereas the logical subject is an experiencer dative DP in the preverbal position.The dative subject in the preverbal position is interpreted as unmarked with respect to the information structure of the sentence: it occupies the clause-initial position without bearing a special pragmatic function.It is neither a topic nor a focus phrase.There is a difference between the unmarked word order in (55), where the dative is realised in the subject position; the sentence in (56a), where the dative is in a focus position (it has focal stress and conveys contrastive information); and the sentences in (56b-c), where it is in a topic position, as shown by the insertion of adverbial material between the dative DP and the verb: The passive sentence in (57b) is characterised by movement of the accusative theme to the nominative subject position, while the agent is realised as a byphrase.Passivization applies when the active sentence has a transitive verb.Intransitive verbs cannot be passivized ( 58b If we assume that a structure is passive if it allows an agentive byphrase, the data in (60) suggest that despite having a nonactive verbal form, quirky subject constructions are not true passives.I treat these as cases of pseudopassivization in the sense that they have a passive form but an active meaning. 16he treatment that I propose for the passive is based on Chomsky (2001).I assume that passive sentences are characterised by a defective little v that is unable to value the accusative case on the theme.When the case of the direct object is not assigned a value, it can enter into an agreement relation with the functional head T. In passives, the direct object moves to SpecTP, where it obtains the nominative case.Agree involves ɸ features and the case of the DP and the EPP feature of T, which is satisfied by raising the direct object to SpecTP.The case of the direct object depends not on V but on the higher nondefective probe T.The agent, which is optional, is realised as an adjunct byphrase, so its case is assigned/checked by the preposition.
In particular, for Albanian passive sentences (such as (57b)), I assume that the morphological nonactive verbal form is the realisation of a voice head, heading a VoiceP projection17 and taking a vP complement.The h-roles for the agent and theme are discharged via merger.The agent is merged in SpecvP in the same way that the subject is merged in active sentences, and the theme is merged in the object position of V.This reflects the uniformity of the theta assignment hypothesis (Baker, 1988(Baker, : 46, 1997: 74: 74): Identical thematic relationships between items are represented by identical structural relationships between these items at the D-structure level.
In passive sentences, the verb raises to v, which in turn moves to voice to incorporate the affixes.The direct object raises to the SpecTP.The agent is adjoined to VoiceP.The derivation I propose is shown in ( 63): (63) In this proposal, the relevant property of passive sentences is their nonactive morphology, which blocks the assignment of the accusative case,18 forcing the direct object to surface as a nominative.This, in turn, blocks the raising of the subject to the SpecTP position (occupied by the object).In passive sentences, the raising of the direct object is correlated with the assignment of the structural case in the landing SpecTP position.The movement of the two arguments, subject and direct object, is driven by case reasons.Despite having a nonactive verbal form motivating the existence of a voice head, quirky subject constructions are structures without byphrases, characterised by only one of the two DPs, the theme, needing a case assigner.The dative case on the experiencer is an inherent case.Evidence that the dative is an inherent case comes from passivization.If we apply the rule of the passivization to a sentence such as the one in (64a), we see that only the direct object can be moved to the subject position (SpecTP), where it obtains the nominative case and controls agreement (64b).The goal, with a dative case, can never become a structural subject (64c).The dative case is not affected by the verbal properties of the clause.In passive sentences, the dative DP retains its case: (64) a.To return to quirky subject constructions, the derivation I propose for these sentences is shown in ( 65), where the verb merges with its internal argument in its complement position.The resulting VP projection merges with the external argument in SpecvP; vP merges with VoiceP, which in turn merges with TP: (65) I also assume that in quirky subject constructions, v is the h-role assigner, as in active and passive sentences.The experiencer in quirky subject constructions is assigned a h-role in SpecvP in the same way that it is assigned to the external argument in active/passive sentences.Cases are only partially checked, as in the passive.
In particular, the case on the logical object/theme is checked by T as in passives because the defective v/V cannot check the accusative case, so it is the probe T that assigns, values and deletes the nominative case of the logical object.T agrees with the theme even if the latter does not occupy the subject position (SpecTP).The agreement between the uninterpretable ɸ features of T and the interpretable ɸ features of the logical object is not local but remote.Long-distance agreement with a higher controller is possible since the weak phase status of v/V makes the object position transparent to the C-T probe.Once the case feature of the theme has been evaluated, the DP is inactive; it cannot undergo further movement, so it is frozen in place.This is the reason why the logical object in quirky subject constructions remains in situ.Verbal agreement is with the logical object marked with the nominative case rather than with the dative DP.The dative DP is not an intervener: it is an inactive goal with the case determined in situ: it is checked within the vP by the verb that assigns it a semantic role.However, it moves to the structural subject position SpecTP, but its move-ment is not driven by the need for case licensing.It is the EPP on T that searches for a DP bearing this feature.As the dative DP is the closest goal, it matches the probe T. Therefore, the EPP on T attracts the highest DP to SpecTP.Even if the dative has its case feature valued, it is accessible for further operations.Its movement is driven by the EPP feature on the functional head T. In quirky subject constructions, it is a DP with an inherent case that satisfies the EPP.
Even if Albanian quirky subject constructions have a dative DP, I do not follow the recent literature in using an applicative projection to host the dative because quirky subject constructions differ from the prepositional dative constructions or double-object constructions found in English Beni i dha Xhonit librin Both constructions in (67), which I assume to be derived from the same underlying structure, are active transitive sentences involving three-place verbs projecting an agent Beni, a theme librin and a goal Xhonit.The subject Beni is generated in SpecvP, but it moves to SpecTP, where it agrees with the finite T and values its case and the EPP property of the T-head.The direct object librin is generated inside the VP, and here obtains an accusative case from V. The goal Xhonit is generated in the specifier of an applicative head.This DP is assigned an inherent dative case by the Applhead.The cases of the three arguments are checked by three different probes: T is responsible for the nominative case on the subject/agent, V is responsible for the accusative case on the direct object/theme, and Appl is responsible for the dative case on the indirect object/goal.Quirky subject constructions, instead, involve only two internal arguments-an experiencer and a theme-in a nonactive construction, i.e., a construction that contains only one active probe, T. In quirky subject constructions, the probe T enters two agree relations: first, it checks and deletes the nominative case of the object; second, it agrees with the dative DP, raising it to SpecTP.The two features associated with T, the EPP feature and the case feature, are checked by two different goals: the dative DP checks the EPP feature, while the nominative DP checks the case feature.Thus, a dative subject and a nominative object enter a feature-checking relation with the same functional head, T. I am assuming a split ɸ feature-checking operation involving a single probe T agreeing with two goals: One goal (the nominative logical object) needs to be checked for case; the other goal (the dative logical subject) needs to be checked only for ɸ features.The head V/v is not a probe because its form is nonactive.The distinction between structural and inherent cases is crucial.

PERSON RESTRICTIONS
As discussed in Section 3, Albanian quirky subject constructions display verbal agreement with the nominative object, not with the dative experiencer.Additionally, with some verbs, this agreement is only partial: it involves the feature [person], which must obligatorily be a third person, not the feature [number], which can be either singular or plural.These structures show the same constraints found in Icelandic and Spanish, where the verb agrees with the nominative object, but agreement is restricted to the third person: firstand second-person objects are blocked from controlling verbal agreement.
A common analysis in terms of Person-Case Constraint (Bonet, 1994) has been proposed by many scholars (Taraldsen, 1995;Boeckx, 2000;Sigurdsson, 2000;Anagnostopoulou, 2005;Béjar and Rezac, 2009).The Person-Case Constraint, generalised in Bonet (1991: 182), states that in combination with a direct object and an indirect object, the direct object must be the third person.This prevents the co-occurrence of a dative DP with firstand second-object DPs.
I assume that the same universal restriction characterises Albanian quirky subject constructions with arguments that are an experiencer and a theme.The PCC comes from the need for two goals (the dative and the nominative DPs) to agree with a single probe, T.
Two questions are raised: 1) How are features represented on datives and accusatives?2) How are features licenced?
There is a long-standing tradition in grammatical analyses of treating the pronominal paradigms of the world's languages as in (68)20 : The paradigm in (68) suggests a system of features that distinguishes first-, second-, and third-person DPs.For example, Nevins (2007) assumed that 1P and 2P are specified positively for people, while 3P is specified negatively.Sigursson (1996Sigursson ( , 2000) ) characterises the third person as [-1p, À2p] or 'no person.'Similarly, Boeckx (2000) assumes that a person corresponds to the first or second person, whereas a third person is [-person].
A detailed analysis of the splitting of ɸ features was proposed by Bejar and Rezac (2009).These authors suggest a system of features that distinguishes first-, second-, and third-person DPs and three possible articulations of the probe: 1) a 'flat' probe, which is just [u3]; b) a partially articulated probe [u3/u2]; and c) a fully articulated probe [u3/u2/u1].For each probe, a DP will be a match for every feature of the probe.If a DP is less specified than a probe, it will match only a subset of the probe's features, leaving an active residue that can agree with another DP.For Albanian quirky subject constructions, I assume a split-feature checking similar to Bejar and Rezac (2009).I assume that in these constructions, the person feature of T is checked separately from the number feature.In particular, I assume that the person feature of T is checked by the dative DP.The idea that the interpretable role associated with the dative can be best characterised in terms of [+person] rather than gender or number is discussed in Boeckx (2000).Anagnostopoulou (2005) also assumed that datives are specified for [person] even when they are the third person, while third-person nominative objects lack [person] features.
Once T has checked the person feature of the dative, T is no longer available to check the person feature of the nominative object.Only the number feature of T is available for checking.Therefore, only DPs that have a number but lack a person are allowed in this configuration.To avoid feature mismatch on T, the object needs to be in the third person since only the third object lacks the person feature.Checking the features involves two steps.First, the dative DP checks the person feature of T. Raising to SpecTP also deletes the EPP feature of T. Since T matches the dative DP only in the person feature, the number feature of T remains intact; therefore, an agree relation holds between the probe T and the more remote nominative goal deleting the number feature of T: (69) Even if the dative argument has certain p specifications, it fails to control verbal agreement, allowing the same probe T to agree with the nominative object.When the nominative argument is the third person, the derivation converges; when it is the first or second person, the derivation crashes.Briefly, the agreement between the nominative object and T arises when the ɸ features of T are checked by two different arguments.This means that ɸ features can be articulated into a subset of features, each of which can agree independently.Therefore, the ɸ features of T are not checked simultaneously; the person is checked separately from the number.The person feature of the verb is checked against the person of the dative, while the number of the verb is checked against the number of the nominative.
Therefore, in Albanian quirky sentences, a dative subject and a nominative object enter a feature-checking relation with the same functional head T. The head V/v is not a probe since it has a nonactive form.
In terms of movement, the dative DP increases first because it is closer to T (it is in the minimal domain of v) than the nominative, which is in the minimal domain of V.In (69), the dative experiencer is the only DP that can move to the TP.The raising of the direct object causes ungrammaticality since the experiencer has a blocking effect on the theme: The logical object can be moved only by a focus operation, which also raises the verb to Focus°: (71) NJË CIGARE i lakmohet Benit In contrast, in quirky subject constructions without person restrictions, all ɸ features (person and number) are checked against the nominative NP.The PCC does not apply to them.In these constructions, the agreement of the verb with the nominative object is complete.

CONCLUSION
In conclusion, I have investigated two types of quirky subject constructions in Albanian; both contain a dative casemarked element that occupies the canonical clausal subject position and a nominative object that appears postverbally and agrees with the verb.In one type, the nominative object and the verb have a person restriction: they can be only third-person singular or plural.This restriction is similar to that of Icelandic and is reminiscent of Bonet's (1994) Person-Case Constraint.The other type of Albanian quirky subject construction has no person restrictions.I have shown that these constructions, despite having a nonactive verbal form, do not coincide with passives, and I have analysed them as pseudopassives.
Regarding the obligatory third-person agreement on the verb, which characterises one type of Albanian quirky subject construction, the approach adopted here is based on Béjar and Rezac's (2009) Cyclic Agree, which involves a single probe, T, that agrees first with the dative DP merged into SpecvP and second with the nominative DP in the VP complement.This happens when unmatched features remain on probe T after agreeing with the first goal.
.DEF NACT.3SG sank.AOR.3SG by storm.DEF (ii) Ky libër lexohet lehtësisht nga studentët this book reads.NACT.PRES.3SGeasily by students.DEF (62) Uniformity of the Theta-Assignment Hypothesis (UTAH) In transitive sentences involving three-place verbs, the subject is nominative, the direct object is accusative, and the indirect object is dative.Albanian has obligatory clitic doubling of indirect objects, meaning that all dative nouns and pronouns in the argument position must be doubled by the corresponding dative clitic: TV is fixed to John by Ben.'The indirect object cannot become the syntactic subject (11a); it can move to the initial sentence position, but it maintains the dative case (11b).The indirect object cannot bear the nominative case: 'To John, it is fixed the TV by Ben.' kujtoj 'remember', harroj 'forget', josh 'attract', lakmoj 'crave'/'covet', teket 'fancy', vesket 'fascinate'/'seduce', and intransitive verbs, such as eci 'walk', fle 'sleep', qesh 'laugh', rri 'stay', and dhimbsem 'take pity'.The examples in (19) illustrate the case of the transitive verb ha 'eat', whereas the examples in (20) contain the intransitive verb fle 'sleep'.As these constructions have a dative NP, clitic doubling of the experiencer is obligatory: The boys want to eat the cake.'/Theboys feel like eating the cake.' The second type of dative subject construction involves verbs such as kujtoj 'remember', dua 'need'/'want', and lyp 'beg'.This type has no person restrictions, as shown by the examples in (25), where the nominative logical objects triggering verb agreement can be first, second, or third person, singular or plural: Ben remembers her only when he needs her.'Fronting of a nominative object is also possible if the logical object raises to a focus position; in this case, the verb must also raise to a [+Focus] head position.In (28), the leftmost element receives a contrastive focus reading: . In such constructions, first and second nominative objects are allowed: They can occupy the subject position in infinitives embedded under an ECM verb(Sigursson, 1992: 5): Ben.DAT him.CL.DAT remember.NACT.3SGhimself.NOM when be.IMPF.3SGfiveyears 'Ben remember himself when he was five years old.'TheAlbanian dative subjects also pass criterion (32b); in questions, they undergo subject-verb inversion: they occur in the postverbal position (40a), showing the same behaviour as canonical subjects (40b) 10 : Dative subjects maintain their case under raising:Dative subjects pass criterion (32e): the subject of the embedded clause can be controlled by the experiencer in the matrix clause 13 : *Duhet diçka t'i hahet Benit 12 10 Even if Albanian is different from Icelandic because it is not a V2 language, in interrogative sentences it has verb-subject inversion (i).The sentences with quirky subjects display the same behaviour.
Juan i lo preocupa la situación del país t i sin desesperar e i John i CL worries the situation of the country t i without despairing e i 'Nobody likes choral music in this house.'For (44d), dative subjects, such as canonical subjects, can be raised from the complement of a raising verb: The ungrammatical examples in (60) show that the experiencer cannot be realised as a byphrase, nor can an extra byphrase be added to these constructions.Bruening (2013) assumed that byphrases are allowed only with passive verbs 15 ; this makes it impossible to add a byphrase to an active VP.The Albanian example in (61) illustrates this impossibility: