A careful observation of data from middle-passives in Asturian reveals that the optionality of the reflexive clitic se in them is evident in contexts where a verb denoting a change of state or location occurs in them. These verbs may convey both an agentive and a non-agentive reading, as shown in the two possible English translations of the examples in (4); in other words, these verbs participate in the causative alternation (Schäfer 2009, inter alia).

The non-agentive reading in these sentences can be enhanced by means of the insertion of a PP like por sí mesmu (‘by itself’), which can only be licensed in non-agentive contexts. Interestingly, for most speakers,3 the presence of such PP necessarily requires the omission of the reflexive pronoun.

On the other hand, the implicitly encoded generic agent in the agentive interpretation makes it possible for these structures to control into a purpose clause (6), which has been traditionally taken as evidence for the presence of implicit arguments in passives (cf. Bhatt & Pancheva 2006). Crucially, in that case, the reflexive pronoun must be present.

Furthermore, when Asturian middle-passive sentences contain predicates expressing activities or accomplishments, i.e. events notionally requiring the participation of an agent or experiencer, the omission of the reflexive becomes ungrammatical (7); besides, these verbs only allow an agentive interpretation, which prevents them from licensing the PP por sí mesmu (‘by itself’) in these contexts, yet they admit a purpose clause.

Considering the data above, it seems appropriate to tie the presence of se in Asturian middle-passives to the interpretation of a generic agent participating in the event. In structural terms, this can be accounted for by arguing that only the agentive version of middle sentences projects a Voice head which would normally introduce an external argument, yet it is passivized and spelled out by the reflexive clitic pronoun se in Asturian (Schäfer 2008; Suárez-Palma 2019; 2020; 2021). This configuration is indeed a generic se-passive construction.

The analysis presented here relies on Cuervo’s (2003) typology of verbalizing heads; this linguist proposes three different types of event introducers/verbalizing heads, each corresponding to a simple eventive structure. Thus, activities (e.g. lleer, ‘to read’) are verbalized by v_DO; verbs of change or happening (e.g. asoceder, ‘to happen’) by v_GO; and finally, states and existentials (e.g. tar cansáu, ‘to be tired’) by v_BE. Moreover, these events may combine with each other giving rise to complex structures such as causatives (e.g. Florinda quemó les fabes, ‘Florinda burnt the beans’), through the combination of an activity verbalizing head and stative one denoting the resulting state (v_DO + v_BE), or inchoatives (e.g. quemaron les fabes, ‘the beans burned’), by means of clustering a dynamic subevent of change and a stative one (v_GO + v_BE). Given this, the derivation of the generic se-passive in (6) is shown in (8).

Because (8a) is an agentive structure entailing a resulting state, the derivation in (8b) must comprise an activity subevent (v_DO) that triggers the change of state process until it reaches a particular end state (v_BE), i.e. the state of being crumbled in this case. The combination of these two verbalizing heads generates a causative configuration. The locus for the agentive reading is the Voice head merging on top of v_DO, which would normally introduce an external argument in the derivation. Here, however, the reflexive pronoun se spells out the Voice head and renders it passive. The complement of v_BE is the root esmigay-, which undergoes head movement until reaching T^o, and the internal argument esti pan sits in its specifier and is probed to SpecTP by T^o to satisfy its EPP feature, thus licensing nominative case. Following Suárez-Palma (2020), I assume that the generic interpretation is possible through the presence of a generic operator Gen in T^o.

The agentless counterpart, on the other hand, does not host a VoiceP –hence the absence of se– and is therefore a generic unaccusative/inchoative structure. Thus, verbs like those in (7), which necessarily subcategorize for an external argument, can never appear without the reflexive in unaccusative contexts like middles; the same applies for the causative variant of the ones in (4). In other words, the presence of the reflexive clitic pronoun in middle sentences in Asturian is not optional, as it has been proposed in the literature thus far. Instead, it is determined by the type of verb entering the derivation, as well as by its idiosyncratic argument and event structures. The derivation for the non-agentive middle in (5) is given in (9).

What (9) shows is another bieventive structure comprising a subevent of change (v_GO) and a stative one (v_BE), which give rise to an inchoative construction. The root esmigay- is verbalized on its way to T^o. Finally, the internal argument raises to SpecTP to check the EPP features in T^o, becoming the sentence’s grammatical subject.5

In this section I have demonstrated that the reflexive and non-reflexive counterparts of Asturian middle constructions containing a change of state verb show structural differences. On the one hand, the se-variant entails an agentive reading which allows the insertion of a purpose clause, whose infinitive is controlled by the sentence’s implicit external argument; I explained this is due to the presence of a passive Voice projection spelled out by se in the derivation. On the other, the se-less variant is non-agentive because it lacks Voice –hence the absence of se–, which makes it possible to license a by itself PP. Next, I discuss additional data concerning Asturian middle sentences and datives which further support my claim that the presence of the reflexive in Asturian middle contexts involves the projection of Voice.

In Suárez-Palma (2020), I develop an analysis of middle constructions in Spanish containing change-of-state predicates; I point out that Spanish middle contexts containing change-of-state verbs allow the insertion of a non-core dative argument which can be interpreted either as unintentional causer of the change of state (10.i), or as affected by the notional object’s resulting state whether it is internally (10.ii) or externally (10.iii) caused.6

Crucially, I show that the first two interpretations, i.e. unintentional causer (11.i) and affected by an internally caused event (11.ii), become unavailable should a purpose clause be inserted in the structure; in such case, the dative can only express affectation by an externally caused change of state (11.iii). The infinitive in such prepositional phrase is controlled by the implicit generic agent in the sentence –instantiated by means of the reflexive pronoun se– and not by the dative, as shown in (11). It is also the participation of such generic agent in the event that rules out the interpretation of the change of state being internally caused.

Following Cuervo (2003), I explain that an affected applicative head7 is responsible for introducing these dative arguments in the derivation. The dative’s different interpretations are structurally determined by the position this functional head occupies in the derivation. Thus, the affected reading arises when Appl –which is spelled out as a dative clitic pronoun– merges below the first subevent and is therefore applied to the notional object’s resulting state (vP_BE). Depending on whether the first subevent is an activity one (v_DO) (12a) –in generic passive contexts– or one of change (v_GO) (12b) –in generic inchoative configurations– the dative may be interpreted as affected by an externally or internally caused change of state, respectively.

Alternatively, Cuervo (2003) points out that the applicative head may merge on top of the first subevent, giving rise to the accidental causer interpretation; however, I show that this is only possible in the generic inchoative variant (13), i.e. where no Voice head fills the position above v_GO; this accounts for the data in (11).

I conclude that there exists a competition between the applicative head and Voice for the position atop the first vP, and adduces to Wood and Marantz’s (2017) notion of i*, an argument-introducing functional head whose spell-out form varies depending on its surrounding environment. Thus, i* would be realized as Voice^o when its complement is v_DO (14a), and as Appl^o when it is v_GO (14b). The causative instances containing Voice –active or passive– and an affected applicative simply start with two i* in the Numeration (15c); the one that merges before is realized as Appl, and the last one to merge becomes Voice.

Interestingly, the data from Asturian in (15) show that this phenomenon is also attested in this language, although slightly different. In Asturian, the accidental causer interpretation is only possible in the absence of se (15b).

The examples in (15) show that both the generic passive and the generic inchoative counterparts of Asturian middles can host an affected applicative, spelled out as the third person dative clitic pronoun -y in this case, and optionally duplicated by a dative DP (a Xuan). While in the generic se-passive the dative can only receive one interpretation, i.e. affected by an externally caused change of state (15a.iii), the generic inchoative allows two readings for this argument, namely accidental causer (15b.i) and affected by an internally caused change of state (15.ii), but rules out that of affected by an externally caused event (15b.iii). As in the Spanish examples above, these data further demonstrate that the reflexive clitic pronoun se in Ibero-Romance languages is linked to the projection of a passive Voice head encoding the participation of an implicit external argument in the event. Because Voice^o fills the position above vP_DO, Appl^o can only merge below it, therefore denoting the affected reading, but never the unintentional causer one. In the se-less variant, on the other hand, the lack of Voice allows the applicative to merge either above or below vP_GO, which accounts for the two possible readings in (15b). In (16), the two possible merging positions of the applicative are synthesized.

Next, I discuss middle contexts containing alternating unaccusative verbs of movement.

The reflexive pronoun –although optional– is occasionally required with certain unaccusative verbs of movement for deictic (17a) or telicity (17b) reasons in languages like Spanish; in Asturian, however, it is banned.

Cuervo (2014) elaborates an analysis of unaccusative verbs of movement in Spanish that alternate between a reflexive and a se-less variant (e.g. caer(se), ‘to fall’; salir(se), ‘to come out’). The key to her proposal is that the non-reflexive variants align with pure unaccusative verbs in that they convey a monoeventive event of change (v_GO) without entailing a resulting state, i.e. no v_BE (18a); on the other hand, the reflexive counterparts do involve a resulting state, and are therefore bieventive (v_GO + v_BE) (18b), just like the inchoative structures discussed above. Because the nature of these verbs is unaccusative, no Voice head merges in either structure.

The reflexive pronoun in (18b) phonetically realizes the phi-features of the entity denoted by the DP displaying the resulting state, which sits in the specifier of v_BE. Cuervo argues that the function of se in these contexts is twofold: (i) to satisfy v_GO’s requirement to have an argument, as well as (ii) to favor the DP’s reading as the undergoer of the process of change and holder of the resulting state.

These two configurations show several structural differences, including telicity: while the se-less variant can be atelic, the one with the reflexive is necessarily telic, as shown in the examples in (19), with the verb caer (‘to fall’).10

Interestingly, the verb cayer (‘to fall’) in Asturian cannot co-occur with the reflexive under any circumstances, yet it admits both the telic and the atelic readings found in (19) for Spanish.

Cuervo (2014) points out that the reflexive and non-reflexive variants of these unaccusative predicates in Spanish differ in terms of the types of DPs they select as arguments; thus, while the se-less variant can take bare NPs as complements (21a), the se-variant cannot (21b), a phenomenon also contemplated by Masullo (1992) and Fernández Soriano (1999a).13 As (21c) shows, the Asturian verb allows both bare NPs and full DPs.

The data in (21c) showing that Asturian unaccusative constructions are less rigid in allowing both NPs and full DPs as internal arguments than their Spanish counterparts,15 together with the fact that both the telic and atelic interpretations in (19) are available in (20), suggest that the two configurations proposed by Cuervo (2014) (cf. (18) above) are available in Asturian. The difference between this language and Spanish is therefore that v_GO receives no phonological content in the former, as I have been pointing out so far. Moreover, these unaccusative predicates can occur in generic middle contexts which must necessarily be agentless, i.e. generic inchoative/unaccusative configurations, as seen in (17) above, repeated in (22).

As inchoative configurations, these constructions license the insertion of a by itself PP.

Moreover, because these constructions lack Voice, they are expected to allow an applicative head to merge above v_GO, and therefore entail a certain degree of agency. The example in (24) shows that this is the case, and that the dative can control into a purpose clause.

In sum, all the examples above demonstrate that v_GO does not have a spell-out form in Asturian inchoative configurations, and that the reflexive pronoun se encodes an external argument in unaccusative structures such as passives or middle-passives.18 This leads us to discuss an apparent exception concerning unaccusative verbs of movement which appears to challenge this claim in the next session.

The GLLA notes that certain verbs expressing movement such as dir (‘to go’), quedar (‘to remain’), or escapar (‘to flee) may occur along with the reflexive pronoun, although “paez preferible prescindir d’él” (ALLA 2001: 354) [‘it seems preferable to dispense with it;’ translation by the author].

In Spanish, the presence of the reflexive with verbs of movement leads to semantic and structural differences. In addition to the differences in telicity mentioned above –i.e. the reflexive version implies perfective aspect, whereas the se-less variant is interpreted as imperfective–, this pronoun becomes obligatory if the meaning to be conveyed is that of vacating a particular location, as shown in (26) (Sánchez López 2002). In other words, the reflexive plays, in a way, the role of a deictic, referring to a point of departure or arrival, whereas the non-pronominal counterpart denotes movement within a continuum, either spatial or temporal (Maldonado 1997).

Similarly, in the example sentences in (18), repeated below as (27) for convenience, the Spanish verb salir is interpreted as ‘to come out’ in the se-less variant (27a), and as ‘to come off’ in the pronominal one (27b), where the reflexive stands for the aforementioned point of departure. The same contrast can be found in Asturian, i.e. while the non-pronominal version allows both readings (27c), the pronominal one is necessarily interpreted as ‘to come off’ (27d).

Notice that the meaning of these verbs is compositional, since the type of DP they select as their internal argument conditions the possible interpretations. In (27c), the flowers are more likely to be interpreted as the entity sprouting from, rather than coming off, a certain place. However, if we think of a particular structure, like a carnival float, which is decorated with flowers that are attached/glued to it and they suddenly come off, (27c) would be an optimal candidate to express such context in Asturian, whereas the reflexive would be required in Spanish. Additionally, when a dative occurs in sentences containing these verbs in Asturian, this argument assumes such deictic role, which translates into certain redundancy if the reflexive pronoun is also present. Note that this is so independently of whether the dative argument is inanimate or not, as show in (28).

In light of these data, I will continue to claim that Asturian, again, does not require the reflexive in (27d) and (28a), in spite of their being inchoative configurations, because v_GO does not have a phonetic form in this language. Moreover, I will assume that the fact that the reflexive is possible for a number of speakers in those contexts is a transfer from Spanish, due its major influence and dominance in the region; this adds on to the list of attested phenomena arising from the contact between these two languages (D’Andrés 2002, inter alia).

This transfer is not exclusive to verbs of movement. In fact, those undergoing the causative alternation can occasionally admit the reflexive pronoun in their agentless variant.

What is interesting about the example in (29a) is the fact that structural ambiguity arises between an inchoative construction and an active causative configuration with an omniscient subject if the by itself PP is omitted. On the other hand, in (29b) the ambiguity emerges between an inchoative structure and a se-passive. It might well be that Asturian speakers choose the pronominal variant in order to disambiguate cases like (29a). In fact, if the theme in (29) occurs preverbally, the sentence becomes unquestionably inchoative, and the reflexive becomes redundant.

Finally, it should also be noted that the opposite is true too, i.e. Asturian Spanish (AsturSp) also shows transfer from Asturian in this regard. Thus, it is possible to find se-less inchoative contexts in AsturSp, such as the ones in (31a-c), which are compared to their respective Standard Spanish counterparts (31a’–c’).20

In this section, I have provided additional evidence that inchoative configurations in Asturian are non-pronominal, and I have proposed that those cases where the reflexive arises in such contexts can be accounted for by adducing to the influence of Spanish, the dominant language in the region. Moreover, I have shown that the transfer is bidirectional, since instances of se-less inchoative sentences are abundant in AsturSp.