Galician Perfective Periphrases among Complex Predicates: Degrees of Grammaticalization and the Possibility of a Perfect Tense

: The so-called per í frasis perfectivas in Galician present the action as concluded or realized. This particular aspectual feature constitutes the common ground for an otherwise heterogeneous set of constructions, ranging from rematar de ‘finish’+ infinitive (e.g., remat ó u de beber ‘(s/he) finished drinking’) to ter ‘have’ + participle (e.g., teñen ido ‘(they) have gone (Rep.)’). This work provides a critical assessment of their syntactic and semantic properties in cases where the participle may not show agreement. This is the case for periphrases built on three auxiliaries: ter , levar , and dar , of which ter + participle stands out as the most grammaticalized one. The case of ter is further investigated in relation to European Portuguese (EP) and Brazilian Portuguese (BP), where ter + participle is considered a fully-fledged perfect tense. Additionally, the use of these periphrases in areas where Spanish is also present is evaluated from a contact perspective.


Introduction
The Galician verb system as presented in descriptive grammars appears to be fully composed of synthetic forms.Table 1 illustrates this point with a prototypical paradigm.
However, it would be inadequate to say that Galician lacks analytic verb forms altogether: these are complex forms in which two verbs contribute a single domain of predication, and Galician indeed has a variety of these periphrases.And while it might be true that this language lacks an analytic tense of the kind that is usually identified in other Romance languages, a perfect with a fully-fledged auxiliary (although see §3 for conflicting evidence), it nevertheless has a variety of periphrases in which a finite verb with various degrees of grammaticalization 1 is accompanied by a non-finite verb (an infinitive, a gerund, or a participle).
From the point of view of linguistic analysis, periphrases are part of the more general domain of complex predication (Svenonius 2008), and it is from that angle that the study will unfold: firstly, I provide a short introduction to complex predication as it is addressed in linguistics more generally; secondly, I move on to examine the Galician perfective periphrases and how they fit into the overall picture ( §2); thirdly, I analyze the (micro)variation in the use of the most grammaticalized periphrasis, ter + participle, establishing a comparison between Galician and varieties of Portuguese (EP, BP); lastly, I give a short overview of the use of these periphrases in Spanish.
In a very broad sense, a complex predicate is a sequence of elements X Y which together serve a predicative function (Svenonius 2008, p. 48).The defining property of a complex predicate is that it involves only one syntactic domain of predication: in other words, it is a monoclausal structure.Monoclausality, sometimes understood as joint predication, is also assumed in the definition of periphrasis in Romance (García Fernández 2006).For a V-V (verb-verb) construction to be monoclausal, only one verb may contribute the lexical information, while the other one contributes the more functional information related to tense and aspect.The prototypical candidates are auxiliary constructions such as the perfect tense in English (John has danced), where there is a functional, inflected verb have which anchors the event in time and does not convey the meaning of possession of have as a lexical verb, and a non-finite verb dance, which contributes the lexical information of a dancing event.Finding reliable tests for monoclausality in these verbal sequences has proven to be a difficult task (García Fernández 2006;Gómez Torrego 1999;Yllera 1999), because the division of labor between the inflected and the uninflected forms is not always so clear-cut.For example, Gómez Torrego (1999) notices that the same combination in Spanish (volver a INF) is more or less monoclausal depending on the semantics of the infinitival form.Compare (1) and (2) below: If volver a INF is a monoclausal structure, the inflected verb should systematically drop the original meaning that it has as a main verb, mainly 'coming back'.This is true of (1), but not of (2).
Another property that has been argued to follow from the monoclausal nature of a V-V combination is its syntactic cohesion as a unit, disallowing the insertion of lexical material in between the two members of the complex.For example: (3) It appears that the only stable distinction that one can posit is between the two ends of the grammaticalization line: on one side, fully lexical, main verbs, and on the other side, fully functional auxiliary verbs.Then the tests work quite well.But how do we account for all the cases in the middle?This puzzle motivated Butt (1995Butt ( , 2003Butt ( , 2010) ) to propose a third natural class of verbs, between auxiliaries and lexical verbs, which she called light verbs: they are defined by being semantically lessened (therefore, "light") with respect to their lexical version, and predicationally dependent on a main verb.They differ from auxiliary verbs, however, in that light verbs are able to modulate the lexical information of the main verb in subtle ways, serving as a sort of Aktionsart-changing device.We can illustrate this with an example from Galician: the periphrasis dar + participle in (5): ( In (5), it is impossible to reconstruct the original meaning of a giving event from the verb dar; instead, it is the participle that establishes the predicative content of an arrival event (see also Pöllänen 2022).And yet, the contribution of dar as (approximately) 'manage to' in ( 5) modulates what would otherwise be a simple arrival event.Patterns like this have been analyzed as instances of "semantic bleaching" (Bybee et al. 1994).
In what follows, we will test the clausal status of some V-V combinations in Galician, mainly among perfective periphrases, establishing a division between constructions where the participle shows morphological agreement with an object, and those in which it does not.Among the non-agreeing cases, we will identify dar as a prototypical light verb in the sense of Butt, contrary to ter, which will be presented as a (more or less) fully-fledged auxiliary construction.Lastly, levar will be shown to constitute a case of a rather functional verb which nevertheless exhibits some of the properties of light verbs.In this way, the work presented here contributes to extending the empirical coverage of theories of complex predication and posits new puzzles for the current classifications (Butt 2010).

The Galician Perfective Periphrases as Complex Predicates
In this section, I will be focusing on the following V-V combinations in Galician, where participial agreement seems optional: The possibility for agreement on the transitive participles has been reported in descriptive grammars of Galician and also in more specialized works such as Rojo (1974, p. 134), but the observation has not yet been further analyzed in a systematic way.This is what this section aims to achieve: a more systematic characterization of the nature of this "optionality".
To that end, cases such as those illustrated in (6-8) were presented to native speakers of Galician from western Asturias (Eo-Navia), who were asked to give an informal impression about the role of agreement.Participants pointed to subtle semantic differences between the non-agreeing version (henceforth, the perfect participle) and the agreeing version (henceforth, the passive participle): in the former case, the sentence focuses on the subject and on the perfective nature of the event, and so, for instance, dado in ( 6) would focus the attention on the fact that the subject is well-travelled and the travel is concluded, whereas dadas would focus the attention on the trips themselves, leaving open the possibility that the subject will travel more.Likewise, xuntado in (8) focuses the attention on the fact that the gathering event took place and it is over, whereas xuntadas leads to a representation of the wooden sticks in the state of being gathered.
Abstracting away from particulars, the subtle differences in meaning that the speakers describe all point to the following generalization: while the passive participle says something about the object, the perfect participle ignores it, and only looks at the subject and/or the event itself. 2  Having established this, there is an interesting parallel with other cases that have been examined more in detail, such as the Spanish auxiliary construction he escrito las cartas 'I have written the letters', with a perfect participle, versus the resultative/adjectival construction tengo escritas las cartas 'I have the letters written', built with a passive participle.The former is taken to be syntactically monoclausal, while the latter is taken to be syntactically biclausal.The differences in interpretation, which are similar to the Galician cases, would follow from such structural difference between the two types.
What are the syntactic tests that allows us to establish monoclausality in these cases?There are mainly four: word order, categorial substitution, question formation, and transitivity (García Fernández 2006, and references therein).
Regarding word order, while it is possible to move the object to an intermediate position in biclausal structures (9), this is not allowed in cases with perfect participles in Spanish (10): The Galician perfect participles behave as the Spanish perfect, disallowing the movement of the internal constituent in the same way: A second difference lies in the possibility of substituting the participle for adjectives or adverbs: while it is possible to do that with passive participles in biclausal structures (12), it is impossible to do the same with perfect participles in monoclausal structures ( 13 The Galician perfect participles disallow that type of question too, just like the haber perfect in (16).Again, I take ter as an example, but the same applies for levar and dar.Regarding transitivity, levar has its own restrictions, having to do with argument structure, which prevent it from combine with intransitive predicates.The main restriction is that levar needs some sort of quantified object in the event structure, either explicitly (as in ( 6)) or as part of a presupposed set (as in the case of a singular chapter which belongs to a book with several other chapters).However, it is important to highlight how this is different from the strict transitivity restriction on biclausal structures: the difference comes from directional intransitive verbs and prepositional verbs.Levar is able to combine with perfect participles from predicates such as ir a 'go to' and participar en 'participate in' (21)(22), in a way that levar in biclausal constructions with passive participles is not ( 23 Summing up, the morphological difference in agreement in (6-8) is the reflection of a syntactic division between clearly monoclausal structures (i.e., those formed with the perfect participle), and resultative/adjectival structures (i.e., those formed with the passive participle).The structural differences lead to the building of different semantic interpretations: in the monoclausal version, the focus is on the event itself, and/or on the subject; in the biclausal version, the focus is on the object (Detges 2000).
Nevertheless, and unlike what happens with the perfect haber in Spanish, in Galician each of these periphrases has its own idiosyncrasies, which does not allow us to classify the inflected verbs ter/levar/dar as fully-fledged auxiliary verbs: The first two, ter + participle and levar + participle, have the meaning of a perfect; that is, they refer to a past event in relation to a current state, just as haber + participle does in (European) Spanish.However, they necessarily denote plural events.Examples (25-26) illustrate this: the event of being in Rome in (25) cannot be a unique event.Likewise, (26) cannot mean that the books were all ordered in a single ordering event, but rather, the books must have been distributed over (at least) two distinct ordering events.The evidence comes from the Galician spoken in Eo-Navia, unless otherwise indicated.In descriptive grammars of Galician, notably Álvarez Blanco and Xove (2002), these constructions are said to imply iteration and/or quantification: according to them, the ter construction denotes an eventuality that occurred in the past more than one time, and the levar construction expresses a quantitative result that may be numerical (in terms of occasions or number of objects) or partial (in the old sense of portional, part of an action or object realized).Among the illustrative examples given by Álvarez Blanco and Xove, we find ( 27), a case of numerical quantification over occasions, and (28), a case of partial quantification that comes from within the object (the pages).Therefore, despite expressing the meaning of a perfect, the Galician ter + participle and levar + participle are conditioned in ways that are unexpected for a completely functional perfect auxiliary, even though they bear the marks of tense and aspect just as a fully-fledged auxiliary.Only in a flexible approach to complex predication can these periphrases find a satisfactory place.On the one hand, they behave syntactically like perfect auxiliaries and unlike light or main verbs; on the other hand, they imposed semantic conditions on the sentences that they form 4 .
As for dar + participle, it is, as we have previously seen, acting as a light verb in the sense of Butt (2010): it does not express a perfect meaning like ter and levar do, an indication that it is structurally operating on events, and below the temporal-aspectual domain of the clause, which I take to be hierarchically higher, as is generally assumed in generative grammar (see Chomsky 1965 et seq.).

Semantic Variation in the Use of ter + Participle: Galician vs. Portuguese
Among the perfective periphrases that have been introduced in the previous section, ter + participle stands out as the most grammaticalized one, exhibiting properties of prototypical auxiliary constructions.In fact, the formally equivalent construction in Portuguese is described as a fully-fledged perfect tense (Mateus et al. 1989).In this section, I examine the extent to which the formal equivalence of ter + participle as it appears in Galician and in Portuguese is reflected in a semantic equivalence as well, finding that the languages differ in two important respects: one has to do with the systematicity to which pluractional meanings are enforced, and the other one has to do with the possibility of expressing counterfactuality.
With respect to the first issue, the ter + participle in Portuguese is generally presented as a tense that necessarily denotes repeated actions, just as it does in Galician.The relevant examples are presented in ( 29) and ( 30), from Schmitt (2001)  The contrast in acceptability between ( 29) and ( 30), according to Schmitt and other people (Giorgi and Pianesi 1997;Squartini 1998, a.o.), is due to the fact that the eventuality denoted by the participle needs to be able to iterate.A sentence built up around an event of someone's death can hardly be conceived as referring to several events of dying, at least in real life.Hence the awkwardness of (30) 6 .
Nevertheless, iterativity in Portuguese seems to only be required when ter is inflected for present tense.Consider (30), repeated here as (31), and ( 32 The use of a verb like morir 'to die' in ( 31) is not readily accepted due to the semantic mismatch between the event of dying, which is necessarily singular, and the requirement for the event to be iterated due to the presence of the present tense.A similar sentence in the past perfect, however, is perfectly acceptable (32): since iterativity is not enforced anymore, there is no mismatch.
Infinitival uses of the perfect tense are also possible with one-occasion readings: ( According to Molsing (2006), the present perfect in BP does not always enforce pluractional readings.She follows Ilari (2001), and adds "reference to durative situations" as a possible reading in cases such as ( 34 (Molsing 2006, p. 135) In this case, it is not that Maria has been happy in Europe several times, but that the duration of Maria's stay in Europe corresponds to the duration of her state of happiness.
In Galician, the idea that ter + participle carries with it a component of repetition/iteration of the eventuality described is assumed by virtually everyone working on Galician grammar (Rojo 1974;Ferreiro 1996;Freixeiro Mato 1998;Álvarez Blanco and Xove 2002).Nevertheless, some have argued that iteration is required only when ter is inflected for the present tense (García Represas 2004;Álvarez Blanco et al. 1995) Examples (36-38) come from formal written texts, such as newspaper articles and novels.In order to tell whether the source (written vs. oral) had an impact on the availability of non-pluractional readings, I carried out a new search under the label /oral/.The following example was found, from a radio program: Although examples such as (36-39) do exist, we cannot rule out the possibility that these carry the semantics of a Spanish present perfect, even if they are expressed in Galician.This is an open question that goes beyond the scope of the present work, and it is an empirical matter: once we start analyzing different varieties across the Galicianspeaking landscape, we might find variation with respect to the obligatoriness of iterated or pluractional readings.
A second point of variation lies in the fact that Portuguese, unlike Galician, allows ter + participle to appear in counterfactual contexts with a modal function: this is illustrated in (40).The original example is included in Squartini's (1998) monograph on verbal periphrases, taken from M. Torga, A criaçao do mondo, 1937mondo, , apud Suter (1984, p. 84), p. 84).To be sure (40) was not just a literary, obsolete use of the perfect, native speakers of Portuguese (informants from Braga and Lisboa) confirm its validity in present-day Portuguese.Summing up, in this section I have introduced two points of variation between Galician and Portuguese, as well as within Portuguese (EP vs. BP), related to the semantic configuration of ter + participle: one is the systematicity of the requirement that the event denoted is iterated (the pluractionality condition), and what we find here is that the condition is not preserved across tenses in Portuguese, but only enforced when ter is inflected for present tense.Within Portuguese, we find that this generalization might not be true of BP, where present tense inflection does not seem to impose the pluractional semantics either.Galician, in principle, differs from Portuguese in that pluractionality is obtained across tenses as well, and is therefore not dependent on the presence of present tense inflection, but the matter is open to debate, based on examples given above.The second point of variation lies in the possibility of using ter + participle in counterfactual contexts; while this is accepted in Portuguese, it does not seem to work in Galician.

The Impact of Galician Periphrases in the Spanish Spoken in AOC 10
The verbal system of the Spanish spoken in areas of contact with Galician differs considerably from the rest of the country (Harre 1991;Rojo 2005): for instance, it completely lacks the analytic perfect tense formed with the auxiliary haber in cases such as he bailado 'I have danced'.Instead, it resorts to synthetic forms, just like Galician does, to express the meanings that most varieties of Spanish express with haber.Hence: Spanish he bailado > Spanish in AOC bailé (Gal.bailei); Spanish había bailado > Spanish in AOC bailara (Gal.bailara).
Furthermore, Spanish in AOC has a set of periphrases which resemble the Galician ones.Here are some examples from the Galician-Spanish bilingual community of Eo-Navia in western Asturias: Based on the evidence just presented, one could hypothesize that the verbal system of Spanish in AOC is the manifestation in Spanish of an underlying Galician system.If this was the case, the Spanish and the Galician forms would share the same syntacticalsemantic properties, linked to a different morphophonological expression in each language.However, I am going to argue that this is not the right analysis for Spanish in AOC, because it fails to account for a series of differences that we can observe between Galician (and Portuguese, for that matter) and Spanish in AOC, at least in the periphrastic domain.My examples deal specifically with the pairs ter/levar + (perfect) participle vs. tener/llevar + (perfect) participle.
The points of discussion have to do with pluractional readings, the aspectual configuration of the participle, and the semantic nature of the subjects.Firstly, unlike what we see in Portuguese, and arguably in Galician (see §3), the pluractional requirement on tener in the Spanish in AOC is independent of tense: the following examples show how the iterativity condition is preserved across tenses, both for past imperfect (44) and for future (45).The sentences are not accepted in single-event contexts, as in the one selling of a house.The interpretation that speakers of Portuguese assign to (46) is that there were several situations in the past where João was blonde or João acted with cruelty.In other words, Portuguese can take predicates that denote stative qualities in principle and coerce them into bounded eventualities that may be iterated.This possibility is open to transitive statives as well, such as conter 'contain' in (47).That is, Galician ter can take a predicate such as 'be the sustenance of', denoting a stative quality, and make it suitable to count as a series of discrete eventualities where the quality holds.This is exactly what Portuguese does, as well.
With respect to ter + participle, in Portuguese we find that this construction only imposes a condition on pluractionality in the present tense, a generalization which does not seem to apply to all of the Portuguese varieties in the world.Both Galician and Spanish in AOC seem to preserve the pluractional meaning across the board, although the Galician case is not as clear-cut, as I have reasoned before in §3.When it comes to Aktionsart sensitivity, we find that ter in both Galician and Portuguese can take any predicate, including states, whereas tener in Spanish (AOC) can only combine with dynamic predicates.Regarding subjects, Spanish in AOC, as opposed to both Galician and Portuguese, imposes semantic conditions of experientiality on its subjects, both with tener and llevar, whereas in the latter languages it is common to have inanimate subjects on these periphrases.Lastly, only Portuguese perfects can be found in counterfactual contexts with a modal function, and in this case Galician patterns with the Spanish spoken in AOC.

Conclusions
This work has provided a critical assessment of the nature of the Galician perfective periphrases in the light of current analyses of complex predicates.The periphrastic constructions were classified according to the degree to which the inflected verb exhibited auxiliary-like properties under a rather flexible understanding of the functional/lexical division in the verbal domain.The analysis allowed us to isolate three periphrases as being qualitatively closer to fully-fledged auxiliary constructions ([ter, levar, dar] +(perfect) participle), and within them, to identify dar as a prototypical light verb, and ter and levar as expressions of a semantically conditioned perfect tense.
On the empirical side, the study has contributed our current understanding of the variation in the use of the most grammaticalized perfective periphrases, especially ter +(perfect) participle, in the context of the Galician-Portuguese family, but also from a perspective of language contact between Galician and Spanish, finding many subtle points of semantic (micro)variation that will hopefully inspire future work in other parts of the grammar.Understanding grammaticalization as the process by which items that carry lexical meaning come to acquire functional meanings (Hopper and Traugott 2003). 2 In this respect, a reviewer points out a series of interesting works on the diachronic relation of resultatives and perfects: e.g., Amaral and Howe (2010); Detges (2000). 3 There is variation within Galician in Eo-Navia with respect to the maintenance of the etymological lateral approximant [l] in initial position in words like levar: while it is present in some areas (notably the south-west part), it has palatalized in others as llevar.We use the former for exposition, since it is the one used more generally in Galician.See Rodríguez-Castellano (1975) for details. 4 An analytical solution is explored in Jardón Pérez ( 2021), but for reasons of space it is not included in the present study. 5 The examples come from EP unless otherwise indicated.In some cases, examples from BP are explicitly included when they are reported to behave differently from the European varieties.

6
A reviewer pointed out that the predicate morrer in Portuguese can, in fact, be iterated with a plural subject, but this would just be a case of subject distributivity, so that there would be multiple (individual) deaths.

7
I thank João Veloso for this example.
8 Corpus de Referencia do Galego Actual (CORGA) [3.2.] http://corpus.cirp.gal/corga/(accessed on 1 February 2024).Centro Ramón Piñeiro para a investigación en humanidades.9 Example (37) does not mean that the sea level has risen between 10 and 15 cm every year (or regularly) throughout the last century, but that the sea level has progressively increased by 10 or 15 cm during the last century.
10 AOC stands for Areas Of Contact (Galician-Spanish).

11
A reviewer suggests that this restriction may have to do with intentionality or volition, as described in Hopper and Thompson (1980).I leave the matter open.
voltas dado *Darei as estacas xuntado to be stated very clearly. . . the Popular Party has won the election' continued in the seminary, you would have succeeded.'
But this test is not very reliable either.Other V-V combinations that qualify as monoclausal in independent tests allow lexical material between the two verbs: ):Again, the Galician perfect participles behave like the Spanish perfect, disallowing substitutions.I illustrate it for ter, but levar and dar yield exactly the same results. of telling whether an analytic verb form is biclausal lies in the possibility of forming a how question about the sentence.As expected, only the resultative/adjectival construction can be asked in this fashion.
The possibility of combining with intransitive predicates is open to the Galician perfect participles as well.The same applies in the case of dar (see (5)): -24): ): ): . To illustrate their point, they mention cases like(35), where ter in its infinitival form refers to a negated event of going (at all).If this is in fact the case, then Galician would pattern with EP in showing a tensedependent kind of pluractionality.A few examples from the online corpus CORGA 8 seem to go even further, suggesting a lack of pluractionality even in the present tense: By this time tomorrow we will have sold the house' Secondly, ter + participle in Portuguese and Galician are not sensitive to the event type denoted by the participle, so that ter may combine with any kind of predicate in its periphrastic use.We can refer to this property as a lack of Aktionsart sensitivity on the side of the inflected verb.Consider the Portuguese sentence in (46):