Three levels of root insertion in Basque intransitive verbs

Intransitive verbs in Basque vary depending on their subject case and auxiliary selection: (i) some of them (considered unaccusatives and inchoatives) always have an absolutive subject and an intransitive auxiliary; (ii) others (prototypical unergatives) show diachronic and dialectal variation (absolutive subject and intransitive auxiliary vs. ergative subject and transitive auxiliary); and (iii) other unergative verbs with which an ergative subject and a transitive auxiliary prevail cross-dialectally. We propose a sublexical structure (Ramchand 2004, 2008) where verbs can be decomposed in three subevents. Unaccusative verbs in Basque are Path predicates, selecting a V of process (VPROCP) that, in the case of telic verbs, takes an adpositional phrase (PP) as complement. Atelic unaccusative verbs and unergative verbs having intransitive morphology are similarly Path predicates, but instead of a result PP, they can select a Rheme of process. In unergative verbs with transitive morphology no V of process is projected, but a little v of initiation (vINIT or vDO). Verbal roots may be inserted at three levels in Basque: vP, VP or PP. The last two are below an AspectualP involved in absolutive case assignment and intransitive auxiliary selection. Thus, depending on where the root is inserted has an effect on the alignment.


Introduction
In Basque, subjects can be marked with ergative or absolutive case depending on the predicate in which they occur: when the verb is bivalent transitive 1 , as in (1), the subject is marked with ergative case (-k), and the object is marked absolutive (-ø). When it is monovalent, the subject can be paralleled to either the object of the transitive verb and get absolutive case -following an ergative case system -, as in (2), or to the subject of a transitive verb and get ergative case (3).
(1) Jone. In an inflected clause with an analytical verb form, the ergative subject always occurs with a transitive auxiliary (*edun or *ezan HAVE) (3), whereas the absolutive subject always goes together with an intransitive auxiliary (izan or *edin BE) (2).
The subject case and auxiliary selection in monovalent verbs have usually been considered a manifestation of different underlying argument structures: those taking ergative subjects are unergative verbs, while those selecting absolutive subjects are unaccusatives (Levin 1983, Salaburu 1992, Oyharçabal 1992. In this paper, we focus on this kind of alignment variation, especially on dialectal data which shows that subject case and auxiliary alternation do not always match the distinction made between unaccusatives and unergatives. Unergative verbs can be of two types in Basque: (i) complex unergative, consisting of a bare noun, an adverb or a PP, plus a light verb 'do' (3); and (ii) simplex unergative, formed by a single verbal word (4). Many simplex unergative verbs have a complex counterpart (3)(4). The complex variant is preferred in southwestern varieties although simplex verbs are also commonly used in central varieties.
(4) a. Irati The alignment in the example (4b) parallels that found in typical unaccusative verbs, like the one in (2). In this paper, we analyze subject case and auxiliary alternation with intransitive verbs and suggest that it is derived because the root lexicalizing the verb has been inserted in different decomposed verbal projections: (i) in v INIT P or v DO P; (ii) in V PROC P or V GO P; or (iii) in P C P. We claim that an Aspectual phrase is sandwiched between vP and VP (Albizu 2001, Ritter & Rosen 2005, Travis 2005, MacDonald 2010 and also in the form of T O in Pesetsky & Torrego 2004) and that it is involved in the intransitive alignment of verbs. Therefore, we assume a sublexical syntactic structure, similar to Ramchand's (2004Ramchand's ( , 2008 First Phase Syntax and also compatible with an account of different flavors of v as in Cuervo (2003) and Folli & Harley (2005). Intransitive predicates whose root is inserted in V PROC P and P C P will have an absolutive subject and BE auxiliary, because Asp necessarily subcategorizes for V PROC . Those intransitive verbs whose root is inserted in vP will surface with transitive morphology, since no AspP is available for them.

Intransitive eventive verbs of Basque 2
Intransitive verb types (mostly taken from Albizu 2009) can be classified in three groups depending on their subject case and auxiliary selection pattern.
According to the General Basque Dictionary (OEH) -a historical corpus -, absolutive subjects were general for these verbs in all varieties prior to the 18th and 19th centuries (except for igo/igon 'go up' and urten/irten 'go out' 7 ). Looking at modern data in the Contemporary Reference Prose (2000--2006), we can see that the ergative marking is the most common in southern texts (although absolutive marking is also accepted and used with some verbs), while the absolutive subject is general in northern texts. The questionnaires that we have conducted confirm this fact, but they interestingly show that ergative subjects are used in meal related verbs and some verbs of manner of motion among young speakers of northeastern varieties.

Group C: always ergative and *edun 'have'
Finally, there is a set of verbs whose subjects are always marked ergative in all varieties and times. Among the simplex verbs, some of them are denominal, like distira N -tu 'glitter', but others are loan words whose root does not exist independently in the language, such as funtziona-tu 'work'. Among them we find non volitional emission verbs (for example, izarniatu 'twinkle', distiratu 'glitter', usaindu 'smell'), non animate activity verbs 8 (funtzionatu 'work', zirkulatu 'circulate'), irakin 'boil' and complex unergative verbs (X + egin 'do') which belong to different semantic classes.

Path predicates
In those verbs belonging to Group A, a salient semantic feature is Path. Talmy (2000) considers Path as another element of a motion event: "The basic motion event consists of one object (the Figure) moving or located with respect to another object (the reference Ground) […]. The Path is the path followed or site occupied by the Figure object with respect to the Ground object" (Talmy 2000: 26). We suggest that Path can also be understood as the transition of a state towards another, as well as the trajectory traversed from a given position towards another. In (5), before the eventuality has taken place, the Figure, Irati, was standing up. She has undergone a transition and she is then seated. Consequently, the Figure has reached the Ground, which is the new posture of being seated. In (6), Irati was not at the same place as the speaker before the eventuality took place. After that, it has reached the reference Ground, which is the deictic place occupied by the speaker.
Not all verbs from Group A denote a completed transition; some of them can have an atelic interpretation such as gradable adjectival verbs (Ramchand 2008: 27) like ilundu 'darken' (7). These verbs can combine with a durative adverbial.
(7) Zerua.ø bost minutuz ilundu zen sky.ABS five minutes.for darken.PRF EXPL.(3SG.ABS). be.PAST (eguzkia berriz agertu zen arte) (until the sun appeared again) 'The sky darkened for five minutes (until the sun appeared again)' The interpretation of this kind of verbs can vary: it is possible to understand that the Figure is undergoing the transition of becoming darker (but not totally dark) (7), or otherwise, that it has finally reached the state of being dark, for which case the transition would be completed. In the former interpretation, the one in (7), ilundu is a predicate of progressive change. We consider both interpretations, the telic and the atelic one, realizations of Path predicates.
We suggest that these two interpretations are the consequence of having two different event structures, which are syntactically realized. In the verbs denoting a completed transition, the interrelation introduced by an adposition is a basic component. According to Hale & Keyser (1993: 71) adpositions denote interrelations where some entity comes to be involved in a relation (the Figure) with another entity (the Ground). In the same way as in the verb etxeratu 'go home', which is visibly derived from an allative phrase (etxe N --ra ALL , literally 'house-to'), as suggested by Oyharçabal 2003, we claim that the telic interpretation of these verbs is constructed from an adpositional phrase. Following Van Riemsdijk & Huybregts (2002), Svenonius (2008) and Ramchand (2008), adpositions can be decomposed into PathP and PlaceP. The Place/Path distinction parallels the contrast between interrelations of central coincidence and of non-central coincidence (Hale 1985 Different interpretations regarding telicity in the eventive domain can be explained in similar terms: verbs denoting a completed transition project a verbal head (Vº) of progression or process (paralleling the meaning of an adposition of terminal coincidence -Pathº -but in an eventive relation) which subcategorizes for an adpositional phrase of central coincidence (PlaceP or P C ) (9a). On the other hand, when the verbs only denote progression, they only project V PROC P (9b).
In (9a) the adposition (which could be null, as in telic deadjectival verbs) is incorporated to the verbal head, in a similar way as suggested for denominal location verbs in Hale & Keyser (1993). In the case of telic verbs like etxeratu 'go home' (10a), both the N and the adposition are incorporated onto the verb through cyclic incorporation. In the atelic interpretation of ilundu 'darken' (10b), the A is incorporated directly to V PROC .
(10) Thus, the V PROC P would be the minimal projection shared by all verbs of Group A.

Root insertion in three levels
We suggest that intransitive verbs can have their roots inserted in three levels of the event structure (three subevents in Ramchand 2004Ramchand , 2008 In the previous section we have argued in favor of the presence of a Path predicate, represented in V PROC , in the verbs of Group A (those which always select an absolutive subject), and we have claimed that the root is incorporated to V PROC from its complement position in atelic verbs -the second level of insertion outlined above -and from the complement position of a P C P, through cyclic incorporation onto the verbal head in telic verbsthe third level of insertion.
In this section, we further propose that in those dialects where the verbs of Group B are realized in the intransitive alignment (absolutive subject and BE auxiliary) the root is also incorporated from complement position onto the verbal head of process (just like in atelic verbs of Group A). Instead of denoting a progressive change, they denote manner of progression (the difference of meaning is given by the categorial status of the complement and the encyclopedic meaning of the roots).
In the formation of these unergative verbs (those of Group B with absolutive subjects and BE auxiliary), an incremental object is incorporated onto the verb. This object can be paralleled to the Rheme of process analyzed in Ramchand (2004Ramchand ( , 2008. According to Ramchand, Rhemes of process are the complements of the process subevent and they unify with the topological properties of the event, giving rise, in some cases, to bounded events. We propose that in the case of Group B unergative verbs of Basque, the predicate is derived by means of the incorporation of these complements onto the verbal head (11). This would be the underlying derivation for the verbs of Group B selecting an absolutive subject and an intransitive auxiliary.
(11) vP Nevertheless, as mentioned in the beginning of this section, this kind of incorporation is not the only process by which an unergative verb can be formed. The verbal root can also be inserted in a higher level, in a complement position of a v of initiation. This is the case of the intransitive verbs of Group C like distiratu 'glitter' or borroka egin 'fight'. In the simplex verb distiratu, the complement root distira is incorporated onto a null v DO (12a). As for complex verbs like borroka egin or distira egin, v DO has its own root inserted and doesn't need its complement to be incorporated (12b).
(12) a. vP b. vP We believe that the structure generated in these derivations (12) cannot be paralleled to transitive constructions, even though the subject ends up getting ergative case in both cases. The reason to believe it is mainly the following: it is still not clear to us whether what is incorporated or just stands in complement position is a √P (funtziona-tu 'work'), an NP (disdira NP egin 'glitter'), an AdvP (hagin-ka AdvP egin 'bite') or a PP (hega-n PP egin 'fly'). In our opinion, this element cannot constitute an argument by itself; it is not a DP. According to Pesetsky & Torrego (2004), case is an uninterpretable T S /T O feature on D. If this would be true, this element would not be able to get absolutive case, like direct objects do in transitive structures. 9 In order to account for the dialectal variation found in verbs of Group B, we suggest that these intransitive verbs have been reanalyzed by some speakers of southwestern varieties in the way of the verbs of Group C. Instead of inserting the root in complement position of V PROC , they introduce it in the complement of v DO . We claim that the use of these two different mechanisms gives rise to the different alignment patterns attested in Group B intransitive verbs. We suggest that the absolutive case assignment and BE auxiliary is mediated by an Aspectual phrase sandwiched between vP and VP. Consequently, the level where the root is inserted (in v DO P or V PROC P) will have an effect on the alignment (see section 5).
The reanalysis of these verbs could have been motivated by a bigger presence of complex unergative verbs in southwestern varieties of Basque, which would be reinforcing the mechanism of root insertion in complement of v DO in intransitive verbs of manner of progression. 10 Verbs of Group C, on the other hand, do not show diachronic or dialectal alignment variation, because, under our view, incorporation into a verb of process is not available for them.

Cognate, hyponymous and target objects
Certain unergative verbs are able to license cognate or hyponymous objects while others cannot. This observation has been previously made in works such as Fernández (1997) and Etxepare (2003). In addition to this, there is an interesting relation between this behaviour and the alignment variation (Etxepare 2003, Berro 2010: the unergative verbs which can take Rhemes of process are those which show diachronic and dialectal subject case and auxiliary variation. Only verbs of Group B can take cognate or hyponymous objects (13) or target objects (14). This result can be explained under our analysis. The unergative verbs which can have Rhemes of process are those where the root is inserted in V PROC P, while those where the root is inserted in v DO P cannot take rhematic objects.
(16) VP As anonymous reviewers suggested, someone can think that those speakers who have reanalyzed the verbs of Group B are not going to be able to have a cognate object construction with these verbs, which is contrary to fact. When an overt specific rhematic object is present in the syntax, a V PROC is going to be necessarily selected, even for southern speakers who have reanalyzed the verbs of Group B. A verb of process is still available for them, since it is also used in other many contexts, such as in intransitive verbs of progressive change or completed change and incremental theme structures. When there is not such a rhematic object, and the verb is intransitive, it is freer to be analyzed in one way or another. 11 Verbs in Group C do not offer this possibility, since they are necessarily inserted in v DO P and not in V PROC P.
The detailed derivation of cognate or hyponymous constructions is hard to understand (16). We can think that both the root which is going to be incorporated onto the V PROC (√dantza 'dance') and the cognate object (dantza bat 'a dance') are generated in complement position of the verb, within the root phrase (√P) (like in Marantz 1997, Cuervo 2003, where roots can introduce complements) but this is an issue that needs to be studied in future.

Case and auxiliary variation
We suggest that the intransitive alignment of eventive intransitive verbs is mediated by an Aspectual Phrase (AspP) that is sandwiched between vP and VP (a similar AspP position is defended in Albizu 2001, Travis 2005, Ritter & Rosen 2005, MacDonald 2010, and also in the form of T O in Pesetsky & Torrego 2004). The subevents represented by V PROC and P C P (or resP) are below AspP and are aspectually active (MacDonald 2010). 12 Without having in the structure at least V PROC , AspP is not going to be present and an intransitive alignment is not going to be obtained. Case and agreement features are going to be checked higher in the functional tree, presumably in TP, so that the ergative case and HAVE auxiliary surfaces. Therefore, depending on where the root is inserted, the predicate can be aligned with 11 Actually, some unaccusative verbs are starting to be used in the transitive alignment among some southwestern speakers. 'A fool stoned two pedestrians' This contrast can be explained considering that the simplex form harrikatu has been generated below AspP, in V PROC P or P C P, whereas the complex harrika egin would be generated in v DO P. transitive (ergative subject and HAVE auxiliary) or intransitive morphology (absolutive subject and BE auxiliary).

(17) vP
The predicates that select V PROC or both V PROC and P C P are going to be aligned in the intransitive way, because they are located below AspP. On the other hand, the intransitive predicates that only project v DO are going to receive transitive morphology from TP, since no AspP is present in the syntax.
Nevertheless, and as an anonymous reviewer notes, intransitive morphology is used in many contexts in Basque, such as in some stative verbs (like izan 'be', egon 'stage level BE' or bizi izan 'live'), impersonal constructions and the progressive. There must be other way to obtain the intransitive alignment (that might include AspP and (i) the use of a V BE parallel in position to V PROC , or/and (ii) a less specified variant of the abstract morpheme in T) but it is still unclear for us. However, since the result is the same in all these contexts, a unitary explanation of case and auxiliary selection would be the most desirable solution. It needs further research.

Conclusion
We have given a unified account of the intransitive verbs selecting absolutive subject and intransitive auxiliary, by claiming that they are all Path predicates and that they are all in a syntactic structure involving a verb of process or progression (V PROC ). We have claimed that in eventive intransitive verbs lexical roots can be inserted in three different subverbal levels: in v DO P, in V PROC P and in P C P. For this aim, we assume a subeventual structure similar to that proposed in Ramchand (2004Ramchand ( , 2008). An Aspectual phrase (AspP) would be present between v DO P and V PROC P, and it would be involved in the intransitive alignment of the predicates. In addition, the location of this aspectual phrase explains the aspectual properties of intransitive verbs: (i) verbs whose root is inserted in P C P are telic verbs; (ii) those inserted in V PROC P are telic or atelic (and have, in the case of progressive verbs, the ability of becoming telic through the use of Rheme of process); and finally, (iii) those in v DO P are atelic.