2.1 Basque verbal (and non-verbal) configurations

Apart from a few verbs like joan ‘go’ and etorri ‘come’,¹ the vast majority of inflected verbs in Basque are analytic, consisting of a verb in a participial form and an auxiliary (izan ‘be’ or edun ‘have’), which includes all inflectional morphology except for aspect. In turn, the participial form of the verbs takes different aspectual markers, each of them giving rise to a different aspectual interpretation: depending on the tense of the auxiliary, perfect or perfective in (1), imperfective in (2) and prospective in (3).

The adjectival participial form that is analyzed in this paper is built on top of the perfect/perfective participle – which will be called bare participle for convenience, following Zabala (1993) –, and consists of an additional morpheme -a (glossed as Pred here) and an optional number agreement morpheme attached when the subject of predication has plural number, as in (4).

Therefore, the (verbal) bare participle and the (adjectival) -a ending participle differ in the use of the -a morpheme and number agreement: the bare participle shows no number agreement with respect to the subject (1), whereas the participle ending in -a takes a -k morpheme corresponding to the plural when the subject of predication is plural (4). Apart from these morphological contrasts, the -a ending participle differs from the bare participle in that the former has a more restricted aspectual interpretation. In an example like (4), the participle is interpreted as a resultative: in the reference time (present), the girls are in the state of being in bed as a result of the prior event of going to bed.²,³ The bare participle in (1), in contrast, can be interpreted as the perfect of result, as an experiential perfect or as a perfect of recent past (Comrie 1976).⁴

Regarding the derivation and the morphological composition of the -a ending participle, someone could think that -a is actually standing for an adjectivizing functional head. However, this is not the position that I adopt here. In fact, the -a morpheme is only necessary when the participle occurs in predicative position (6). When it modifies a noun, in attributive position, as in (5), the participle can be bare when the noun phrase is headed by the indefinite determiner bat ‘a’ (5a). The -a morpheme has exactly the same distribution in individual level adjectives like gorri ‘red’.

If -a was an adjectivizing morpheme, we would expect it to appear also when the participle is in attributive position (5a), and we would not expect it to occur with root-derived adjectives like gorri ‘red’. On the basis of that, I will claim that -a is the exponent of a Pred(icative) head (Baker 2003), extending the analysis of -a made by Eguren (2012) for non-derived adjectives.⁵

Thus, verbal and adjectival participles are similar in form: without taking -a into account, the participle occurring in perfect or perfective contexts is morphophonologically the same as that occurring in attributive and predicative positions. Nevertheless, they behave syntactically very distinctly. Unlike bare participles, -a ending participles occupy positions that are usually filled by adjectives: only -a ending participles are compatible with the locative copula egon ‘be’ (7a), can serve as complement of AP-selecting verbs (irudi, eman ‘seem, look like’) (7bc), and can be modified by the degree adverb oso ‘very’ (7c). Bare participles are not allowed in these contexts (8).

Furthermore, another context where we find a contrast between bare participles and -a ending participles are interrogatives. In wh-questions formed from clauses involving a bare participle, the participle is fronted together with the inflected element – the auxiliary in this analysis – to the position immediately following the wh-word (9a). In contrast, in interrogative clauses built from -a ending participles, only the inflected element – the copula – is fronted to that position (10b). An example where the -a participle is fronted together with the copula is ungrammatical (10a). Similarly, the example where the bare participle is left behind is also ungrammatical (9b), at least in southern varieties of Basque.⁶

In this respect too, -a ending participles behave like root-derived adjectives:

These distributional facts show that even though bare participles and -a participles have the same form, they have different syntax. More specifically, I consider that bare participles (8) are verbal, whereas -a ending participles are adjectival (7) and that to be the reason for their different distribution. Consequently, I will claim that when the latter occurs in predicative position, like in (6a) and (7a), the inflected element (izan ‘be’ or egon ‘[locative] be’) is really a copula, and not an auxiliary (Ortiz de Urbina & Uribe-Etxebarria 1991; Hualde et al. 1994; Artiagoitia 1995; de Rijk 2008). In contrast, the inflected element that combines with the bare participle is an auxiliary. This explains the fact that only izan ‘be’, and not egon ‘[locative] be’, is compatible with the bare participle (8a): izan ‘be’ can be an auxiliary in Basque, but egon is not. The locative egon only combines with non-verbal predicates: adjectival or adpositional phrases. It accepts to have -a ending participles as complement because -a ending participles are adjectival.⁷

A further clarification note is in order here: unlike in other languages, adjectival participles are not passive in Basque (Eguzkitza 1981; Ortiz de Urbina & Uribe-Etxebarria 1991; Artiagoitia 1995), given that they are acceptable with ergative arguments interpreted as initiators of the event, just like verbal participles (see section 3.4). Nevertheless, as I will show in section 3.3, the syntactic distribution of ergative subjects is not the same in verbal and adjectival participles, as their syntactic arrangement in the clausal architecture is different.

2.2 Different types of adjectival participles

Adjectival participles have been claimed to be of a hybrid category, with properties belonging both to verbs and to adjectives. They behave like verbs in that they (usually) denote an event and in that they can be, in certain cases, modified like verbs, but they also resemble adjectives in that they occur in positions that are usually reserved to adjectives, such as in attributive or complement predicate position and in complement position of AP-selecting verbs like seem.

On the other hand, as has been noted in the literature, adjectival participles do not form a homogeneous group. Starting from Wasow (1977), deverbal -ed forms in English have been claimed to be formed at different levels of the language architecture. Depending on whether an event is implied or not, and on the acceptability of different event-related and subject-oriented modifiers, adjectival participles have been argued to be of different sizes, with different structural compositions, particularly involving different verbal layers (e.g. Anagnostopoulou 2003; Embick 2004; Alexiadou & Anagnostopoulou 2008; Bruening 2014; Doron 2014). Kratzer (2000) classified adjectival participles into resultant state and target state participles, and Embick (2004), in turn, differentiated between resultative and stative adjectival participles. In this paper, we will focus on these two typologies, and I will add a third aspectual class – particularly to Embick’s two-way typology –, which corresponds to the participles that are interpreted experientially.

Kratzer (2000) classifies phrasal participles into two different semantic classes: target state participles and resultant state participles (Parsons 1990). Target state participles denote a state that is reversible and that can be modified in German by the adverb immer noch ‘still’.⁸ In contrast, resultant state participles denote a state that is not reversible; a state that is the result of a culminated event. Anagnostopoulou (2003) and Alexiadou & Anagnostopoulou (2008) show that in Greek only resultant state participles – and not target state participles – involve the projection of VOICE. This asymmetry can also be observed in Basque, as I will show in section 3.

On the other hand, Embick (2004) classified English passive adjectival participles in two groups, apart from the verbal participle (called eventive): stative and resultative.

Stative participles denote a state, like a non-derived adjective, whereas the resultative denotes a state that is the result of a previous event, where this event is grammatically represented. In the former, the participle is syntactically structured with a stativizing Aspectual head (Asp_S) projected directly on top of an acategorial Root, whereas in the latter, the Root is verbalized (in a verbal phrase, vP) before being selected by Asp_R.⁹,¹⁰

There are different diagnostics to tell apart stative participles from resultatives ones. One involves the use of adverbs that modify the manner in which the event has taken place (Kratzer 1994). Stative participles do not allow this kind of modification (13a), whereas resultatives do (14a). Another one is the ability to occur as complement of verbs of creation like build, create or make (Embick 2004). Resultative participles are not acceptable in this environment (14b), since they imply a previous event and this is incompatible with the presence of the verb of creation. Stative adjectival participles, in contrast, are perfectly grammatical in this context (13b).

Thus, even if the same root (√OPEN) is used in (13ab) and (14ab), the contrasts found suggest that it is being used in different configurations, and this is supported by the participial morphology (-ed) only found in the resultative and not in the stative adjective. In other cases, however, the same participial form is used in both (e.g. closed) (15).

Kratzer’s and Embick’s classifications do not have a one-to-one correspondence. In fact, as explained by Alexiadou et al. (2014), Kratzer’s (2000) target state participles are not necessarily stative. They can have event implications and therefore, they are argued to be built on top of vP (Anagnostopoulou 2003; Alexiadou & Anagnostopoulou 2008; Alexiadou et al. 2014; Alexiaodu et al. 2015). Similarly, Kratzer’s resultant state participles do not correspond to Embick’s resultative participles, since target state participles can also be resultative. The relation of the two typologies is illustrated in Table 1.

Participles in Basque can be of all these types. Roughly speaking, bare participles correspond to Embick’s eventive participles, although they are not passive in Basque (see section 3.4). As for adjectival participles, -a ending ones are generally resultative, but in the case of some verbs, they can also have stative interpretation. The examples in (16) show that the -a ending participle ireki-a ‘open/opened’ is compatible with a verb of creation like eraiki ‘build’ (16a), and accepts to be modified by manner adverbs (16b). Therefore, ireki-a ‘open, opened’ can be both stative and resultative.

This is not the case in all -a ending participles, though. For those participles that have a non-participial stative counterpart (stative predicates ending in -(r)ik), like the synonym of ireki-a, zabaldu-a ‘opened’, the -a form must be interpreted as resultative.

Both zabal-ik ‘open’ and zabal-du-a ‘opened’ involve the same root: √ZABAL ‘wide, open’. The former, ending in the suffix -(r)ik, is compatible with the verb of creation eraiki ‘build’. The latter, the -a ending participle zabal-du-a ‘opened’, sounds weird in that context. As expected, the former does not accept to be modified by manner prepositional phrases (18a), whereas the latter accepts them naturally (18b). These contrasts indicate that, in this case, the -a ending participle is really a resultative, and not a stative participle.

As for the target state and resultant state difference, in section 3 I will show that -a ending participles can be exponents of both types of participles. There is a clear contrast though: only resultant state participles allow temporal and spatial modification of the event, and the presence of dative and ergative arguments related to the event underlying the participle. Additionally, in section 4 I will argue that we can add a further aspectual type to the four-way distinction of adjectival participles in Basque: those that are interpreted experientially.

3.1 Acceptability of event-related and subject-oriented modifiers

Basque adjectival participles are compatible with event-related and subject-oriented modifiers, a fact that indicates, firstly, that the event is actually grammatically encoded in the structure.

As noted by Kratzer (1994) for German adjectival participles, the compatibility of adjectival phrases with manner-denoting modifiers supports the presence of the verbal phrase below the stativizing Adjectival head. As we can see in Basque, the acceptance of the prepositional phrases arretarik gabe ‘sloppily’, lit. ‘without any attention’, and kontu handiz ‘carefully’, lit. ‘with much care’, shows that the events denoted by the participles orraztua ‘combed’ and konpondua ‘fixed’ are grammatically encoded. I take this to be evidence for the projection of v below the Adjectivizing layer.

On the other hand, an aspect that has generated much discussion in the study of adjectival participles deals with the presence of VOICE, the head introducing the external argument (Kratzer 1994; 1996). Kratzer (1994) claimed that adjectival participles are different from verbal ones precisely in the projection of this head: unlike verbal passive participles, adjectival ones are not VOICEP-derived, and thus, an implicit external argument, e.g. an agent, is totally absent from them. Nevertheless, later works questioned this claim, firstly, for Greek (Anagnostopoulou 2003; Alexiadou & Anagnostopoulou 2008), but more recent studies have also argued that VOICEP can be involved in English and German adjectival participles (e.g. Alexiadou et al. 2014; Alexiadou et al. 2015 for German; McIntyre 2013; Bruening 2014 and Alexiadou et al. 2014 for English).

In the literature, the absence of the external argument in adjectival participles has been supported using several diagnostics, such as the inability to control into purpose clauses, and the absence of the disjoint reference effect. Nevertheless, the lack of the disjoint reference effect as a test to prove the absence of an implicit external argument in German (Kratzer 1994) has been called into question in McIntyre (2013), Alexiadou et al. (2014) and Alexiadou et al. (2015). On the other hand, the inability of adjectival passives to control into purpose clauses in German and English has also been counter-evidenced in McIntyre (2013) and Alexiadou et al. (2014).

Leaving aside the tests of control and the disjoint reference effect (which also yield positive results for the presence of VOICEP in Basque), Basque adjectival participles are compatible with modifiers that rely on the presence of an agent, such as the prepositional phrase kontu handiz ‘carefully’, lit. ‘with great care’ (20) and the adverb nahita ‘on purpose’ (21).

The projection of VOICE is also evidenced by the presence of ergative subjects and instrumental phrases that are related to the event underlying the participle (and its subject).

The idiot sailor and the can opener are participants involved in the underlying event. Particularly, the ergative subject is interpreted as the initiator of the event, a fact which shows that it is introduced in the specifier of VOICEP, below the adj(ectivizing) layer. The instrumental phrase is a subject-oriented manner adverb.

It must be noted that an active VOICE may not be projected in all kinds of adjectival participles. More specifically, as noted in section 2.2, Anagnostopoulou (2003), Alexiadou & Anagnostopoulou (2008), Alexiadou et al. (2015) argued that target state participles do not include the VOICEP projection. In relation to this, Basque data also support the absence of an active VOICEP projection in target state participles: subject-oriented modifiers and ergative subjects are not accepted in combination with the modifier oraindik ‘still’.¹²

The adverb oraindik ‘still’ would be grammatical in the absence of subject-oriented modifiers and ergative subjects. Therefore, I conclude that such modifiers and arguments are only acceptable in resultant state participles, and not in target state participles.¹³ As will be shown in the next subsection, resultant state participles are also compatible with other event-modifiers like temporal and spatial PPs.

3.2 Compatibility with temporal and spatial modification

Another test that is used to explore the structural composition of adjectival participles deals with the temporal and spatial location of the event underlying in the participle. If participles are compatible with temporal and spatial adverbs or prepositional phrases that locate the event at a time and at a place different from those of the state (also denoted by the participle), the event must have been actually instantiated, in terms of Gehrke (2011; 2015). This means that the event denoted by the participle has taken place in the actual world; it is an event token rather than an event kind.

A modifier like recently or a time-frame adverb like three days ago can be used to test the nature of the event underlying the participle. Gehrke (2011: 246) shows that in German adjectival participles, the adverb recently can only modify the state and not the event (25), and that the prepositional phrase three days ago is not acceptable at all (26) (von Stechow 1998; Rapp 1996; 1997).

As can be seen, only a temporal prepositional phrase that modifies the state, like seit drei Tagen ‘since three days ago’, is compatible with German adjectival participles, since German does not accept modification of the event underlying the participle that is not relevant to characterize the state (McIntyre 2015).¹⁴ On the contrary, such prepositional phrases are acceptable with Greek adjectival participles ending in -menos (Alexiadou et al. 2014; Alexiadou et al. 2015).

Contrasting with German, and patterning with Greek, Basque allows temporal and spatial modification of the event underlying the participle.

Nevertheless, temporal and spatial modification is not totally free. It seems to be better with activity verbs – those having a ‘job is done’ interpretation (in the sense of Kratzer 2000; see also Irmer & Mueller-Reichau 2018 for recent discussion) – and with achievements and change of state predicates that yield resultant states (geldituak ‘had done an appointment’, konpondua ‘fixed’), rather than those creating target states. Thus, participles built on verbs of change state that denote target states, e.g. zabaldua ‘opened’, do not sound very natural with temporal/spatial modifiers of the event.

Thus, we conclude that participles denoting non-reversible situations allow more freely locating the event at a time and at a place different from those of the state. According to Gehrke (2015), event instantiation takes place when the verb is embedded under further verbal projections like ASP and T. Given that Basque resultant state participles are compatible with temporal and spatial modification,¹⁵ I propose an analysis where this kind of participles are functionally rich: before the adjectivization takes place, the Root is embedded under vP, VOICEP and ASPP, the latter with the meaning of anteriority, as it has been argued for Greek (Alexiadou et al. 2014; Alexiadou et al. 2015).

3.3 Compatibility with dative indirect objects and ergative subjects

Another distributional feature that shows that adjectival resultant state participles denote an instantiated event (rather than an event kind) is that they are compatible with unrestricted dative objects and ergative subjects, just like inflected analytic verbs. Dative is the case usually assigned to indirect objects in Basque, whereas ergative is born by agents, causers and external arguments in general.¹⁶

Nevertheless, the presence of these kinds of DP arguments in the adjectival participial configuration is syntactically much more restricted than in common inflected verbs (see Ortiz de Urbina & Uribe-Etxebarria 1991). As a matter of fact, unlike in configurations consisting of a bare participle and an auxiliary (34) (35), the dative and the ergative arguments do not agree with the inflected element (the copula) in the adjectival participial configuration (32) (33). In perfect configurations such as (34) (35), the dative and the ergative must agree with the inflected element (the auxiliary); otherwise the examples are ungrammatical.

Another remarkable property of dative and ergative arguments in adjectival participial configurations is that their position is fixed: they must occur after the absolutive theme and before the participle. If placed in other positions, like before the absolutive theme or after the participle, the result is ungrammatical.¹⁷

This is not the case in perfect analytic clauses. For instance, given that Basque has quite a free word order within clauses, the dative and ergative arguments can occur in a variety of different positions.

Following Ortiz de Urbina & Uribe-Etxebarria (1991), I claim that these distributional differences arise because -a ending participles occur in biclausal configurations. The separation into two clause-domains explains the facts. The embedded clause is headed by the tenseless participle, and the matrix consists of the copula that, in turn, takes the participial predicate as complement. As agreement and scrambling in Basque are clause-bound, the arguments belonging to the verb embedded under the adjectival participle stay in the participial domain: the arguments do not move out of it and do not agree with the copula above. The only argument introduced at the matrix clause is the absolutive theme, which necessarily agrees with the copula, as can be seen in the examples above. This is also why it is possible to select an intransitive copula (egon or izan ‘be’), rather than a transitive one (eduki or *edun ‘have’), when the verb embedded in the participle is transitive and has an ergative argument. In the perfect analytic configuration consisting of the bare participle, the auxiliary must be transitive (*edun ‘have’) if the verb is transitive and occurs with an ergative subject, as in (35).

Contrary to the adjectival participial configuration, the perfect analytic configuration consisting of the bare participle is monoclausal. The inflected element in that configuration is an auxiliary, rather than a copula, and as a consequence, the dative and the ergative argument show free word order and must necessarily agree with the auxiliary.

3.4 Lack of restrictions in dative and ergative arguments

Gehrke (2011 and subsequent work) observes that by-phrases in German adjectival participles are possible in two scenarios: first, if they belong to the consequent state (40a) rather than to the underlying event; or second, in the case they are event-related, if they have an impact on the underlying event that is still visible in the consequent state (40b). In the latter, only certain types of by-phrases are accepted. For instance, Gehrke & Marco (2014) observe that the complements of por-phrases in Spanish adjectival participles are generally weak or non-referential nominals such as indefinites, bare nominals and weak and generic nominals and, among other things, they cannot have wide scope, like in (41). According to Gehrke & Marco (2014), the reason for these restrictions is that these kinds of por-phrases modify an event kind, rather than an event token.

As shown in section 3.3, modification is quite unrestricted in Basque. I have shown that ergative arguments that belong to the event – and that are not necessarily relevant in the consequent state – are accepted in resultant adjectival participial configurations. Furthermore, the nature of the ergative argument (as well as of the dative argument) is unrestricted: we can have strongly referential nominals in this position, such as proper names (32) (33), pronouns, demonstratives and regular definites (42) (43).

Therefore, Basque differs from German and Spanish in this aspect too, and again patterns more with Greek, which actually accepts apo-phrases with strong nominals, such as proper nouns.

As mentioned in section 2.1, it must be noted that Basque adjectival participles are not passive, as their external argument is introduced as a regular ergative DP instead of as a by-phrase.¹⁸

The languages that have been mentioned in this paper so far – English, German, Greek and Spanish –, and which have passive adjectival participles, are all nominative. As Basque differs from them in being ergative and in having transitive (no passive) adjectival participles, it is tempting to think that these two properties are somehow related. For instance, the lack of passives may not be restricted to adjectival participles but might be related to the ergative nature cross-linguistically, given that the passive configuration is less common in ergative languages (Kazenin 2001: 926; apud Polinsky 2017). In relation to this, there has been much discussion about the syntactic nature of passivization in ergative configurations (Otsuka 2000), as these processes do not seem to be totally productive, are sometimes carried out by different morphological strategies within the same language or are argued to involve a stativizer morpheme. In particular, Basque seems to lack passives altogether (see in this respect footnote #18). If passivization is not available generally in Basque, it follows that it will not be available in adjectival participles either. This means that the VOICE head in adjectival participles is not necessarily syntactically defective and that the adjectivizing or the Aspectual head involved does not always existentially quantify over the external argument (see section 5). In turn, these two aspects may be related to the fact that external arguments are licensed in their generating position in ergative languages, as it is assumed in the inherent view of the ergative case (Legate 2002; 2008; Massam 2002; Aldridge 2004; Laka 2006; Woolford 2006; and many others). In ergative languages like Basque, external arguments do not need tense in order to be case-valued, contrary to what happens in nominative languages. In adjectival participles of nominative languages, agent arguments must be introduced resorting to by-phrases. Basque simply does not need to do that, as agents can be introduced and licensed as in regular verbal configurations. In any case, the relation between transitive adjectival participles and ergativity is an issue that needs to be studied further.

3.5 Interim conclusion

In previous sections, we have seen that Basque adjectival participles allow event-related modification, as they are compatible with manner prepositional phrases. Furthermore, in the case of resultant state participles, subject oriented adverbs and prepositional phrases, temporal and spatial modifiers, as well as dative and ergative case-marked participants of the event can also be included, with certain syntactic restrictions that make them different from typical dative and ergative arguments introduced in inflected analytic verbs. In particular, dative and ergative DPs belonging to the event underlying the adjectival participle do not agree with the copula (the inflected element), and their position is fixed, i.e. they must remain between the absolutive theme and the adjectival participle. I have argued, following Ortiz de Urbina & Uribe-Etxebarria (1991) that these restrictions are due to the fact that adjectival participles occur in a bi-clausal structure: the inflected element is a copula that takes the participial predicate as its complement. In this way, predicative configurations involving adjectival participles differ from perfect analytic configurations with bare participles, given that the latter are monoclausal, with the inflected element as an auxiliary, rather than a copula.

Finally, I have also shown that the dative and the ergative DPs can be strongly referential, contrary to what happens with by-phrases in other languages such as German or Spanish.

All this evidence shows that in Basque adjectival resultant state participles, the event underlying is really an instantiated event, rather than an event kind. Gehrke (2011) and following works (Gehrke & Marco 2014; Alexiadou et al. 2014; Alexiadou et al. 2015; Gehrke 2015) claim that adjectival participles in languages like Spanish and German are interpreted as consequent states of event kinds because in these languages, the verb does not get further embedded into verbal projections, but is directly adjectivized after vP (or VOICEP in the case resultant state participles). Gehrke (2015) generalizes to verbs the analysis of nominals (Carlson 1977; Zamparelli 1995; Chierchia 1998; Dayal 2004 among others); both nouns and verbs enter the derivation as predicates of kinds and get instantiated when they are embedded under functional projections; Number in nouns and Aspect or T in verbs. At this level, they obtain the interpretation of entity tokens or event tokens (with spatiotemporal location). In languages like German and Spanish, resultant state participles are not further embedded under Aspect or T (they are directly adjectivized), and as a consequence, the interpretation is of an event kind. This means that the event underlying the participle cannot be temporally or spatially modified – as it lacks spatiotemporal manifestation – and that the acceptable modifiers either describe the consequent state or, otherwise, are interpreted as kind-related; modifiers that create a new subkind of the event by narrowing the event kind, and thus, having an impact on the consequent state.

Aleixadou et al. (2014) and Alexiadou et al. (2015) show that Greek -menos adjectival participles allow non-restricted event-modification and argue that, in Greek, the adjectivization takes place after an Asp(ectual) operator has been projected. Thanks to ASP, the event underlying the participle is instantiated and can be freely modified. Following the same line of analysis, I will claim that in Basque adjectival resultant state participles too, an ASP head is projected before the adjectivizing head (see section 5). Nevertheless, in Basque, Asp must be able also to accommodate the experiential interpretation, given that Basque adjectival participles can also be interpreted experientially under certain circumstances (see section 4).

4.1 Not four, but five kinds: Experiential adjectival participles in Basque

In section 3, I showed that -a ending resultant state participles pattern with Greek -menos adjectival participles in that they imply a previous event that is actually instantiated. In that resultative interpretation, the participle is generally derived from a telic predicate and its consequent state holds at the relevant evaluation time. Therefore, it is similar to the Perfect of Result, as portrayed in Comrie (1976), Iatridou, Anagnostopoulou & Izvorski (2001/2003) and other works.

But Basque -a participles are even more flexible than Greek ones (Alexiadou et al. 2014) as they can have, under certain circumstances, an experiential meaning. This specific interpretation can be observed in the examples below, where the participles are built on the predicates ikusi ‘see’ and egon ‘[locative] be’. In the case of the transitive ikusi ‘see’, the absolutive argument cross-referenced in the copula – the theme of predication – is linked to the subject of the transitive verb embedded in the participle, rather than to the object (de Rijk 2008).

The experiential interpretation is generally defined as a reading in which the subject has a certain experience related to the fact that the predicate has held at least once in the interval spreading from sometime in the past until the reference time (Comrie 1976; Iatridou et al. 2001/2003; Pancheva 2003; Mittwoch 2008). It describes an eventuality that more generally occurred at a time before the reference time and that the subject has accumulated some experience as a consequence. Therefore, the experiential interpretation contrasts with the resultative one in that the outcome of the event does not necessarily hold at the relevant evaluation time. It is usually triggered by iterative adverbials like many times or more than once (46) and plural direct objects that yield series of independent eventualities, like pelikula asko ‘many films’ in (45).

The fact that non-verbal participles can have an experiential interpretation was already observed in the diachronic study carried out by Krajewska (2012; 2013a; b), a work that analyses Basque texts written from the 17^th to the 20^th century. According to her, the experiential interpretation is favored when the situation has held many times and, as a consequence, the subject is considered an experienced person due to what she/he has lived. In both (45) and (46), the subject is ascribed some qualities as a consequence of a past repeated action. Nevertheless, the experiential interpretation is not restricted to this context, as it can also only be obtained when the event is not repeated: in the example in (45) the experiential reading remains even if behin baino gehiagotan ‘more than once’ is replaced by behin bakarrik ‘only once’ or even if the sentence is put in the negative (thus denying the eventualities at stake).¹⁹

This experiential interpretation of -a participles is generally linked to the presence of cognitive (e.g. learn, realize), perception (e.g. see, hear) and communication (e.g. say) predicates where the eventuality affects the subject, rather than the object or, alternatively, to the presence of stative and unergative verbs (e.g. be, travel), that is, predicates that do not yield a good result state. The examples above involve a perception verb (ikusi ‘see’) and a stative verb (egon ‘[locative] be’). Furthermore, in (45), the theme of predication is co-indexed with the subject of ikusi ‘see’, rather than with the object. This is a necessary condition for adjectival participles derived from transitive verbs to be interpreted experientially. In fact, the experiential interpretation is restricted to this particular scenario: on the one hand, the event underlying the participle must be of the appropriate type and, on the other, in the case of transitive verbs, the subject of the copula must correspond to the subject of the embedded predicate. If these conditions are not met, the resultative interpretation arises.²⁰ As an example, it is hard to combine a participle derived from a change of state verb like hondatu ‘break’, where the subject of the copula is co-indexed to the object of the embedded verb, with a frequency adverb like askotan ‘many times’ that would force an experiential meaning (Mittwoch 2008).

According to Krajewska (2012; 2013a; b), the resultative reading is highly predominant with change of state verbs, which historically are also the most frequent verbs in this configuration. Even if no change of frequency has been observed in the classes of verbs used during the last centuries, Krajewska notes that the experiential interpretation increases from the 16^th to the 20^th century; according to her, it starts to appear with a certain frequency in the late 19^th century.

Like with resultant state participles, experiential -a participles are compatible with temporal and spatial modification of the event underlying the participle, a fact which shows that experiential participles of this sort also involve an instantiated event.

The experiential is one of the possible interpretations of the Perfect, as pointed out by Comrie (1976). The fact that -a ending participles can have a particular type of experiential interpretation might suggest that this configuration is really a perfect construction, involving a verbal participle and an auxiliary. Krajewska (2012; 2013a; b) actually claims that this kind of participles are in the middle of a grammaticalization process towards a new perfect. She points out that, in the presence of cognitive, perception and communication predicates, as well as stative and atelic unergatives predicates, the experiential interpretation has become more common in the last centuries. According to her, even though it is still not a full-fledged perfect, given that the frequency of the latter verb classes has not increased in comparison to change of state verbs, these participles are acquiring properties usually attributed to the perfect. Nevertheless, I consider that, at this moment, the configuration involving the -a participle is semantically and syntactically very different from the perfect. In fact, there are many distributional facts that point in this direction. Firstly, the experiential interpretation is restricted to certain verb classes. Change of state verbs (e.g. lose) do not give rise to that interpretation. As a consequence, adding a following clause that reverses the consequent state results in a contradiction (50).

In contrast, in the perfect configuration consisting of the bare participle, the experiential reading can be obtained with any verb, and such a continuation is acceptable (51). Secondly, not all the readings of the perfect are available in the -a ending participle. For instance, the universal perfect (or perfect of persistent past, as in Comrie 1976) cannot be obtained. In the perfect consisting of the bare participle, stative verbs can give rise to an interpretation where the state denoted by the predicate started in the past and still holds at the reference time (52) (Iatridou et al. 2001/2003; Pancheva 2003). Such an interpretation is unacceptable in -a ending participles, and as a consequence, since-type modifiers like ablative postpositional phrases with temporal meaning are ungrammatical in that context (53).

Thirdly, as I have shown in section 3, -a ending resultant state participles are compatible with temporal modifiers that locate the event at a time different from that of the consequent state (54). This type of modification is not possible in perfect clauses consisting of the bare participle in Basque (Artiagoitia 1995) (55).²¹

The incompatibility of the perfect with temporal adverbs that locate the event in the past is known as the “present perfect puzzle” (Klein 1992; Pancheva & von Stechow 2004), a feature observed in the perfect configuration in languages such as English and Spanish. Interestingly enough, -a ending participles in Basque – and also in Greek, as pointed out in Alexiadou et al. (2015: 182, footnote # 27) – contrast with participles used in the perfect configuration in allowing this kind of modification.

In the next sections, I will continue dealing with the nature of -a ending participles, focusing mainly on the experiential reading. Particularly, I will discuss the category and the syntactic distribution of experiential -a participles, and I will argue that they are adjectival.

4.2 Are experiential -a participles verbal?

A possible analysis of the experiential -a participle and the resultative -a participle would posit structural and categorial differences between the two classes of participles, considering experientials as verbal participles and resultatives as adjectival. McFadden & Alexiadou (2006; 2010) adopted such an analysis when accounting for the diachronic evolution of the auxiliary alternation in English, i.e. the replacement of have over be in Early English. At that stage of the language, some unaccusatives were still used with be auxiliary, but they started to show alternation, and finally be was totally replaced by have by the end of the 19^th century. Apparently both configurations started some time before the history of Old English as resultatives. But, the [have + past participle] configuration, they argue, widened its territory and was finally reanalyzed as a perfect in the intervening centuries; it obtained interpretations associated to the perfect, among others, the experiential. In contrast, the [be + past participle] configuration stayed as a stative resultative, which finally disappeared in favor of the perfect configuration.

McFadden & Alexiadou (2006; 2010) propose a different syntax and semantics for each participial configuration. On the one hand, the [have + past participle] configuration yielding the experiential interpretation is argued to be like the Modern English perfect. The inflected element -have- is an auxiliary which is located at an aspectual Perf(ect) head below T (von Stechow 1998; 1999; Iatridou et al. 2001/2003; Pancheva 2003; Pancheva & von Stechow 2004) and which accommodates all the perfect interpretations. The head Perf, in turn, selects for ASPP. They consider that Perf may involve either anteriority semantics, as in Klein’s (1992; 1994) theory, or, alternatively, it may introduce a time span that extends into the past and to which T specifies the ending point (the Extended Now Theory, von Stechow 1999). As they note, the Extended Now Theory would be superior in that it can also yield the universal interpretation of the perfect, unlike Klein’s approach.

On the other hand, they argue that the [be + past participle] is a copular construction built around a stative resultative participle. The verb is headed by Asp_R (Embick 2004, see section 2.2) that, semantically, introduces a state that is the result of an event (particularly, a target state; Parsons 1990). ASPP is, in turn, the complement of the copular verb -be-, a copula just like those occurring with non-verbal predicates. According to them, this structure lacks the temporal/aspectual material denoting anteriority, as the Perf head is not projected.

As can be seen, the two structures differ in terms of their clausal configuration and the category of the participle. In the resultative, the participle occurs in a position usually occupied by non-verbal predicates and it is selected by a copula (57). The experiential have configuration, in contrast, is monoclausal, as the participle is the complement of a Perfect auxiliary (56). Therefore, the participle is in this case verbal. This is not the position that I am taking in this paper to account for the experiential -a participle in Basque. In fact, as I will show in the next section, -a experiential participles have a similar syntactic distribution as that of -a resultative participles.

4.3 Experiential -a participles are really adjectival

Even though they have a different interpretation, -a ending experiential participles have a similar syntactic distribution. In predicative position, they can be combined with the intransitive copulas izan ‘be’ and egon ‘[locative] be’. This is significant especially when the participle is built on a transitive verb (58) (59), given that transitive verbs only combine with transitive auxiliaries (e.g. *edun ‘have’) in Basque. Additionally, egon ‘[locative] be’ is a copula that takes non-verbal predicates as complements; not an auxiliary. These two aspects support the view that -a participles, also with experiential interpretation, are non-verbal predicates, and therefore they can occur in constructions involving intransitive copulas.

Additionally, like resultative -a ending participles, -a ending experiential participles are also grammatical in complement position of adjective-selecting verbs.

On the other hand, if -a ending experiential participles were verbal, and the configuration was a full-fledged perfect, it would be more extended and not so restricted to certain verb classes. Additionally, if it were a perfect, it would also allow the universal interpretation with statives, but this is not possible, as explained in section 4.1. All in all, I argue that we have to consider a fifth type of adjectival participles in Basque (see Table 2): experiential adjectival participles.

The resultative interpretation is predominant in adjectival participles, but as commented in section 2.2, some of them can be interpreted purely statively, without event implications. Furthermore, as shown in this section, when the embedded verb does not yield a good result and/or when the theme of predication is co-indexed with the subject of the embedded transitive verb, the participle is interpreted experientially.

5.1 Stative participles and target state participles

As explained in section 2.2, stative adjectival participles denote states that are not the result of a prior event (Embick 2004). In turn, target state participles refer to states that may or may not be the result of a prior event, and that are reversible. I argue that, in both of them, a stativizing Asp_S/R head has been projected. I follow Kratzer (2000) in considering that Asp_S/R is a stativizer that retrieves (target) state properties from its complement. These state properties are accessed through a state argument that is present in the complement of the Asp head. This is why the Roots or the verbs that combine with Asp_S/R need to have a stative component. In the case of target state participles with event implications, the predicate must be able to decompose into an eventive and a stative component (Kratzer 2000). In this way, we can account for the fact that target state participles are much more restrictive in terms of the type of verb they can involve than resultant state participles. For instance, participles built on activity verbs like erosi ‘buy’ are not compatible with the adverb oraindik ‘still’, even though the event conveyed by erosi is reversible.

I follow Embick (2004) in claiming that the difference between participles denoting an event and those that do not lies in the projection of verbal layers below ASP_S/R. As can be seen in (63) and (64), in statives, ASP_S generally combines with an acategorial root (63), whereas in target state participles, ASP_R merges to VOICEP which in turn involves vP (64). As shown in section 3, target state participles under the scope of oraindik ‘still’ are incompatible with subject-oriented modification or ergative subjects interpreted as initiators of the event underlying the participle. Therefore, I follow Anagnostopoulou (2003), Alexiadou & Anagnostopoulou (2008) and Alexiadou et al. (2015) in claiming that target state participles do not involve a semantically active VOICE head.²⁴ Furthermore, given that target state participles are not compatible with temporal or spatial modification, I argue that in target state participles the event is not instantiated before being adjectivized. As a consequence, the event remains in the kind domain and lacks spatiotemporal manifestation. The category changing must take place before the projection of PRED (externalized as the -a ending). It could be thought to be carried out by ASP_S/R, in which case ASP_S/R can be also depicted as an ADJ head (Bruening 2014) or a little a (Marantz 2001; 2007; Embick 2010) with that particular semantics.

ASP_R selects for a semantically and syntactically defective VOICE (like the expletive non-active VOICE proposed in Alexiadou et al. 2015) (VOICE_{{Ø, Ø}}). Thus, target state participles can only involve a kind of VOICE that introduces neither implicit nor explicit external arguments and does not accept subject-oriented modifiers that would adjoin to a semantically active VOICE projection. On the other hand, v is responsible for providing event-semantics (the change interpretation) and it also introduces an internal argument of type e, x, to which a lambda abstractor has been adjoined (see Bruening 2014). Following the proposal of adjectival passives and the ADJ head in Bruening (2014), I claim that ASP_R attracts the internal argument of vP, a null operator, to its specifier. In this way, the stative property denoted by ASPP is predicated of an open individual.²⁵

In turn, ASPP/aP is selected by Pred (a predicational head) (Baker 2003), and above PredP, the copular verb is projected. According to Baker (2003), adjectives need Pred in order to be combined with a specifier (subject) argument. It is a function of type <<s,t>, <e,<s,t>>>; a function which takes a situation as input and returns a relation between an entity and that situation. Thus, Pred has the function of relating the stative property denoted by ASPP (an adjectival projection) with a referent DP. By Functional Application, the DP introduced in [spec, Pred] binds the free variable in ASPP, and it is, therefore, interpreted both as the holder of the state introduced by ASP and as the undergoer of the event denoted by vP.

5.2 Resultant state participles and experiential participles

As shown in section 3 and 4, in resultant state participles and experiential participles, the event is instantiated. It can be located at a time and at a place different from the state denoted by the participle. As a consequence, I argue that they involve an aspectual operator, particularly, a head that conveys anteriority (Klein’s 1992; 1994 PERFECT aspect; Kratzer 2000; and see also Alexiadou et al. 2014; Alexiadou et al. 2015 for Greek): basically, it situates the situation time (the running time of the event) before the topic time. It selects a property of an eventuality and returns a property of times that is true of any time that follows the running time of the event.

Unlike the ASP_R head of target state participles, ASP_PERF selects for a semantically active VOICE head (VOICE_{{λx, }}). As a consequence, this configuration is compatible with subject-oriented modifiers. On the other hand, the external argument introduced by VOICE can be either explicit or implicit. If VOICE is syntactically active ((VOICE_{{λx, D}}), an argument DP will be introduced in its specifier. If, in contrast, it is syntactically defective (VOICE_{{λx, Ø}}) no argument will be introduced in specifier position and the ASP_PERF head will existentially quantify over the λx introduced by VOICE.

Crucially, in this configuration the projection of ASP_PERF occurs below the adjectival head a, meaning that the event is instantiated before being stativized and adjectivized. Thus, temporal and spatial modification of the event is also allowed. As in target state participles, the adjectival phrase (aP) is the complement of the head PRED spelled out by the -a morpheme, and PREDP is, in turn, the complement of a copular verb.

The semantics of this structure can be explained as follows. vP and VOICEP introduce an event or a series of events related to certain arguments, one of them introduced as an open argument (an argument of type e, which could be an internal or an external argument, to which a lambda abstractor has been adjoined) (see Bruening 2014). This null operator will be then forced to move to the specifier of ASP_PERF, as proposed previously for ASP_R. If it is v the head introducing such argument in its specifier, ASPP will be predicated of the internal argument of the verb (68). In contrast, if the open argument is introduced in [spec, VOICE_{{λx, D}}] –only possible when VOICE is syntactically active– ASP will be predicated of the external argument of the verb (69).²⁶

ASP_PERF has the function of introducing a time interval and relating it to the running time of the event(s) denoted by VOICEP. Particularly, it introduces an interval that follows the running time of the event(s). The adjectivizer, a, introduces a property that holds at that time interval, and finally, PRED relates the property in question with a referent DP (as in target state participles). This DP binds the open argument within ASPP by Functional Application. As can be seen, there are, at least, two different arguments – the subject of predication, introduced in the specifier of PRED, and the open argument internal to VOICEP/vP –, linked by means of variable binding.

As for the aspectual interpretation, the operator ASP_PERF accommodates both the resultant state and the experiential interpretation, given that both denote an interval that comes after the time in which the event holds. I argue that the interpretative difference is determined by the combination of particular vP semantics, the generating position of the open argument and ASP_PERF. If the vP involves a stative verb (70), or an open argument in [spec, VOICE] (71), the participle will be interpreted as an experiential, as a last resort to make the vP/VOICEP compatible with anteriority semantics and the predicative “stative” nature.

In (71), the vP involves an overt DP in object position, and this argument is different from the subject introduced in the specifier of PRED. The specifier of PRED binds, in this case, the open argument introduced by VOICE, the initiator argument and the subject of the transitive embedded verb. Therefore, the property denoted by the adjectival participle is predicated over the subject initiating the (repeated) action, and it is therefore interpreted as an experiential. This would be the structure for the -a ending experiential participle built on the verb ikusi ‘see’, as in examples (45), (48) and (59).

In contrast, if the v introduces the open argument and the verb denotes a change of some sort (so that the open argument is interpreted as the undergoer of the change), the projection of ASP_PERF and PRED will yield a resultant state interpretation, a property predicated about an argument that undergoes a change and which holds at a time that follows the culminated event of change. The argument DP introduced by PRED in its specifier position, in this case, binds the free variable in ASPP, which in this case, is interpreted as the undergoer of the change.

The property denoted by the adjectival participle is predicated about the argument undergoing the change and it is, consequently, interpreted as a resultative. If VOICE is syntactically active (VOICE_{{λx, D}}), a DP will also be introduced in [spec, VOICE], given that ergative initiator arguments can also be found in resultant state participles. Regarding the case assignment of initiator arguments, I will assume that the DP in these cases gets inherent case from VOICE. Thus, I follow the line of analysis developed in Massam (2002), Woolford (2006), Legate (2002; 2008), Aldridge (2004) and many others where it is claimed that ergative case is assigned by the head introducing the external argument (VOICE or little v; VOICE in this analysis).²⁷

Crucially, the configuration illustrated in (67) (and developed in (68) (69), (70), (71) and (72)) does not give rise to the universal reading, specifically due to the semantics of ASP_PERF. In the universal reading, the predicate holds at a time starting from some point in the past until the relevant evaluation time. This aspectual interpretation is incompatible with the meaning of anteriority contributed by ASP_PERF.

The fact that both possible interpretations – the resultant state reading and the experiential one – arise in a configuration involving the same aspectual head is also supported by the fact that they evolved diachronically at the same time.²⁸ According to Krajewska (2012; 2013a; b), the rise in the frequency of experientials goes together with the rise of resultatives that allow temporal/spatial modification (called perfect in her works). At some point in the 19^th century, the stativizing aspectual head turned into an aspectual operator denoting anteriority. Contrary to ASP_S/R, ASP_PERF instantiates the event, and it is able to accommodate the experiential interpretation. Nevertheless, the category of the participle remains the same as in target state participles or stative participles: all of them behave like adjectives, as shown in section 2.2 and 4.²⁹