On the Nature of Verbal Non-Local Doubling in Patagonian Spanish

: The main objective in this study is to describe and offer an account of verbal non-local doubling in Patagonian Spanish (PatSp), an understudied non-standard variety of Spanish in Argentina. We focus on data in which there are duplicated verbs surrounding an XP that bears the nuclear accent of the phrase (XP NA ). First, our analysis describes the prosodic, semantic, and morphosyntactic behaviour of the data gathered. Second, we present the problems and challenges that doubling phenomena in PatSp pose for approaches that have tried to explain similar data in other Spanish varieties and other languages, such as the copy theory or prosodic cloning. Third, this work explores a biclausal analysis of verbal non-local doubling in PatSp in which each duplicate originates in a different clause, CP1 and CP2. In this approach, duplicated verbs (V1 and V2, according to their linear distribution) are not derivationally related. We also argue that the XP NA moves to the left periphery of CP2. This movement would account for the three typical traits of verbal duplication in PatSp: the mandatory adjacency between the nuclear accent and V2, the non-locality between verbal duplicates, and the semantic value of mirativity.


Introduction
Non-local doubling can be considered a crosslinguistic phenomenon in which one element seems to occur twice in the same sentence, as shown in the underlined elements in (1)-( 6).
(1) European Portuguese (Martins 2007, p. 81 As can be observed, in non-local doubling, an element (a verbal or nominal phrase) is duplicated with some linguistic material in between. 1 Most scholars recognise that this phenomenon is associated with particular syntactic, prosodic, and semantic-pragmatic behaviours.In addition, several proposals agree on the fact that the duplicated elements are the materialisation of copies of the same syntactic object.From this perspective, non-local doubling involves the extraction or escape of an element, followed by a remnant movement of ΣP (Saab 2017) or TP (Cvejanov and Druetta 2020), creating the structural conditions for preventing the deletion of one of the copies.Both movements are shown in ( 8 Cases of non-local doubling, then, would be considered instances of multiple copy spell-outs (Nunes 2004), which would give empirical support to the copy theory of movement (see Martins 2007;Jokilehto 2016;Saab 2017;inter alia).
In the current study, we analyse verbal non-local doubling in a non-standard Spanish variety spoken in Patagonia, Argentina, that we refer to as Patagonian Spanish (henceforth, PatSp); see Acuña and Menegotto (1996) and Virkel (2004) for a general description of this variety.PatSp has barely been described from a generative perspective (see Menegotto 2006).In particular, verbal non-local doubling has only been addressed by Silva Garcés (2019). 2

Methodology
The data presented in the current study have been taken from different field trips to the Southern Region of the Province of Río Negro, Argentina, particularly to the Somuncurá area.This is a volcanic rock plateau of approximately 20,000 square kilometres in Río Negro and Chubut provinces in Patagonia, Argentina (41°-43°S and 66°-68°W).The altitude ranges from 500 to 1600 metres above sea level.The highest region corresponds to the volcanic plateau, where living conditions are hostile and there are very few inhabitants.The land is mainly arid, with scarce rains along the year and high temperatures in summer and low ones in winter.The lowest areas have more water reservoirs, allowing for better life conditions.This plateau represents one of the areas with the lowest population density in the country, with inhabitants living in smallholdings several kilometres apart.It is possible to reach most houses by special vehicles through winding and steep gravel roads, but in some areas, the only means of transport is horses.In the inner parts of the plateau, there is no electricity supply or telephone signal.Most inhabitants have radios powered by batteries, which constitute their main source of information (Masera 1998).The environmental features of this region probably explain the lack of population masses.
The data analysed in this study were taken from 19 interviews conducted with people who were born and raised in the Southern Region.The authors of this study conducted 13 of these interviews during different field trips.The remaining six interviews were conducted with local women as part of the development of a community project aimed at retrieving and revalorising life stories by women.In 2017 and 2018, this project was promoted by social workers from hospitals of the cities of Los Menucos, Maquinchao, and Jacobacci, and teachers and students from the region also participated.One of its outcomes is the book Mujeres del viento.Historias de vida de mujeres de la Línea Sur de Río Negro (Women of the wind.Life stories of women from the Línea Sur of Río Negro). 3During the arranged visits, speakers were asked to participate in interviews in which they were free to talk about life in the plateau, their families, childhood, or personal anecdotes.In relation to their mother tongue, all 19 interviewees reported to speak Spanish as their only first language.However, in some cases, they claimed to have parents or grandparents who spoke Mapuzungun as a first language.This means that our participants might have been exposed to some other languages apart from Spanish during their childhood.Another aboriginal language that had been spoken in the plateau was Günün a iajüch, which belonged to the Günün a küne people (Malvestitti 2009).This language was not familiar to our informants, according to their reports.Participants' ages range from 35 to 80. Their occupations consist mainly of rural activities related to agriculture and raising animals such as goats, cows, and horses.Some of them have attended rural primary schools for some years, and most of them have not had any formal education at all.Given that most inhabitants, especially the older ones, do not know how to write, PatSp is mainly a spoken language.Because of the very low population density, we believe that the number of participants in this study is significant.In addition, due to the fact that PatSp speakers have not been influenced by standard Spanish, our data represent a variety that is far from being conditioned by normative restrictions.
All interviews were conducted in participants' places with a portable battery-powered recorder, obtaining WAV files of 24 bit and 96 kHz.Out of the 20 h corpus, 68 instances of non-local doubling were registered in 75 intonational phrases.These instances were analysed perceptually and acoustically by means of Praat (Boersma and Weenink 2023).The phonological analysis is based on autosegmental-metrical (AM) theory (Ladd 2008;Pierrehumbert 1980), following the conventions of Tones and Break Indices in order to transcribe and analyse intonation.In short, this notation employs two levels of pitch accents and boundary tones: H (high) and L (low).Boundary tones can be of two types: they can refer to an intonational phrase (IP) with a level-4 break index (H% and L%) or to an intermediate phrase (ip) with a level-3 break index (H-and L-).For the Spanish intonation, a third pitch level has been proposed in the literature, M (mid) (Beckman et al. 2002;Estebas Vilaplana and Prieto 2008).This level represents a fundamental frequency that is at a midpoint in a speaker's pitch height, which seems to be phonologically contrastive in Spanish.In our study, the contributions for Rioplatense Spanish (spoken around the River Plate and in central Argentina; henceforth RSp) by Gabriel et al. (2010Gabriel et al. ( , 2013) ) and Labastía (2018) have been considered as a starting point in the analysis of PatSp.
The ungrammatical sequences and other data constructed ad hoc for the current research have been corroborated by native speakers of the studied Spanish variety.

Data Description
The main objective of this section is to provide a detailed description of the doubling phenomenon in PatSp.First, we offer a prosodic characterisation of the data in our corpus.Second, we offer an account of the morphosyntactic behaviours, and, third, we show our data from a semantic point of view.In order to systematise the information, an outline of this three-angled description is provided.

Prosody
In relation to intonation and phrasing in PatSp, and from an impressionistic point of view, this variety of Spanish has been described in the literature as having a particular intonational pattern that can be easily imitated (Malvestitti 1993, p. 138).This intonation has been equated to a particular "melody" that differs from other regions in the country (Muñoz and Musci 2007, p. 32;Stell 1987, p. 81;Virkel 2004, p. 185).According to the perceptual analysis by Acuña and Menegotto (1996, p. 251), PatSp can be associated with Chilean speakers who, in turn, have been in contact with Mapuzungun, whose specific pronunciation features can be found in some vowels and consonants and in some intonation patterns.Given that the prosodic studies on PatSp are merely perceptual, the analysis in the current study means a significant contribution to this variety of Spanish in this respect.
As mentioned before, we have analysed the intonation and phrasing of verbal nonlocal doubling in PatSp following the AM theory.In general terms, the great majority of sequences with non-local doubling have a falling intonation, L%, which is related to the fact that these constructions occur in statements with concluding information; only a small percentage of sentences (5 percent) have been produced with a rising intonation, indicating that the speakers have not finished conveying their message (cfr.Labastía 2018 for more details on the interface between prosody and information in Spanish).
To begin with a more exhaustive prosodic analysis, the registered prosodic patterns of RSp found in Gabriel et al. (2010) and Labastía (2018) were considered as a first attempt to describe PatSp.However, some intonational contours encountered in the perceptual and acoustic analysis of our data did not match the inventoried intonation in the consulted bibliography.Specifically, the pitch accent L+H* is not found in nuclear position in RSp when there is a falling intonation, but we propose such a contour for PatSp because L+H* seems to be representative of the melody in the plateau.Our decision is partly based also on the fact that the contour H+L* L-, which is typical of RSp, is found only once in our data.This could suggest that L+H* L-and H+L* L-are phonologically contrastive in PatSp, but further research is necessary.Figure 1 shows the total number of instances of non-local doubling with their corresponding prosodic analysis.As Figure 1 shows, the most typical intonational contour in PatSp is /(L+H* L-)(L%)/.In this notation, there is one IP signalled by slashes and two ip represented by round brackets.In other words, the two ip make up one IP.The first ip, (L+H* L-), is composed of the pitch accent L+H* followed by the boundary tone L-.The second ip does not have pitch accents due to deaccentuation (i.e., there is a lack of pitch accents), and it also ends with a falling boundary tone, L%.The division between the two ip is not signalled by an actual short cease of phonation, as could be expected in a level-3 break, but by an abrupt decrease in the fundamental frequency that is represented in our proposal by L-4 in the first boundary tone.This is illustrated in the spectrographic evidence in Figure 2. The example in Figure 2 clearly shows that the first ip, La Corona saben de ir mucha GENTE 'Many people usually go to the Corona!', contains the first duplicate and the nuclear accent (NA) on GENTE, indicated with small capital letters.This ip is produced with the following nuclear tonal configuration: L+H* L-.The second duplicate, saben de ir '(they) usually go', lacks a prominently accented syllable and ends in a falling intonation, L%.These contours can be seen as fully manifested because there are post-tonic syllables.When there is no relevant linguistic material after the tonic syllable, we still implement the same prosodic annotation because the perceptual impression remains the same.The example in Figure 3 shows a case in which the contour /(L+H* L-)(L%)/ cannot be fully realised due to the fact that there is less material after the tonic syllable. 5In relation to phrasing, our data indicate that the second duplicate categorically occurs in an ip that is adjacent to the NA.In other words, these two elements are never disrupted with a level-4 break, i.e., they do not form two independent IP.This enables us to state that, in the data of PatSp, the NA and the second duplicate should appear in the same IP.This is why a case such as ( 12) is ungrammatical, because these elements are separated with a level-4 break (signalled with slashes).( 12 The examples in (13) demonstrate, then, that speakers can insert pauses before the NA, but they can never do so after it due to the adjacency requirement mentioned above.We can conclude that the second duplicate does not target an independent IP.On the contrary, this element is embedded within a major IP that includes the NA.
In synthesis, from a prosodic point of view, non-local doubling in PatSp shows the following behaviour: • rising pitch accent L+H*; • one IP with two ip; • the first ip has the contour L+H*L-, while the second ip does not have pitch accents due to deaccentuation and ends with a falling boundary tone, L%; • unlike the first duplicate, the second duplicate and the NA are never disrupted with a level-4 break (i.e., two independent IP are not possible).
So far, the structure of non-local doubling in PatSp is shown in ( 14). ( In the following section, we offer a description of this structure in morphosyntactic terms.

Sintactic Behaviours
Non-local verbal duplicates in PatSp show particular syntactic behaviours, which are described below.

Non-Adjacency between V1 and V2
In verbal doubling from PatSp, duplicates are not adjacent.This means that constructions are anomalous if V1 and V2 are contiguous, as (15b)  In addition to this, there is another condition for the above-mentioned lack of adjacency: the XP NA is always between V1 and V2.This is illustrated when the constituents in (15a) are altered.For instance, even when duplicates are not adjacent, the sequences in ( 16) remain anomalous because the constituent between V1 and V2 does not bear the NA. ( 16 As can be observed, our data show that the syntactic function of the XP between V1 and V2 is not relevant: it can be an adjunct (16a), a subject (17a), or a direct object (18a).What matters is that the XP carries the NA.

XP(s) on the Right of V2
In non-local verbal doublings of RSp, V2 is necessarily placed on the right margin of the sequence.In Saab's (2013, p. 240) words, "another relevant property of capicúas 6 is that V2 should always occur in final position.This happens with all types of constituents".The following data illustrate this behaviour.
(19) RSp (Saab 2013, his examples (35, 37, and  In (22a), we can observe that in PatSp, the duplicated elements can display the same root and non-identical morphological features because the duplicated verbs do not have the same tense inflection.In (22b), the same set of syntactic-semantic features (1PL, imperfective and habituality) can be identified, but it is differently materialised: while V1 is a periphrastic form, V2 is a synthetic one.
In addition, V1 and V2 can show lexical differences, as the following data prove.The fact that there are morphological and even lexical differences between V1 and V2 is controversial.Can we still refer to data such as ( 22) and ( 23) as instances of doublings?To what extent can V1 and V2 be different?To answer these questions, the contexts in which these data were obtained should be considered.In this respect, (22a) was produced in a conversation about crops in the area of Somuncurá, Patagonia.Given that this landscape is dry and rains and watercourses are scarce (see Masera 1998 and Section 2), the interviewer asked how fields were irrigated in the past.ahí In (25B), V1 refers to an action in the past, while V2 corresponds to an action in the present.In this particular context, then, the tense in V2 implies a clarification or even a correction with respect to the temporal information in V1.In other words, in the past, there was a stream, but actually it still exists today.
In addition to temporal information, we have found instances of verbal doublings in which V1 and V2 differ in terms of person, as in ( 26 Again, we claim that the features of 1PL in V2 rectify the information in V1.A possible paraphrase of ( 26) would be to use the discourse marker mejor dicho 7 'rather', as in ( 27). ( 27) Traje tres percas esa vuelta.Mejor dicho, trajimos (porque éramos dos personas) 'I brought three perches that time.Or rather, we brought (because we were two guys)' In cases of non-local doublings with different roots, V2 also has an explanatory discursive value.The example in (23a) can be paraphrased as shown in ( 28).
(28) Entraban hasta las culebras, es decir, llegaban ahí 'Even the snakes used to get inside, I mean, arrived there!' In cases such as (23a), an argumental resctriction applies: duplicated verbs select the same number of arguments and assign the same thematic roles.For instance, in (29), the verbs entrar and llegar are monadic and assign the role of theme to their only one argument.

→ theme
If the duplicated verbs did not assign the same argumental information, such as entrar 'get inside ' and cazar 'catch' in (30), the resulting duplicated construction would not be possible.
( As discussed above, V1 and V2 can be identical or they can present some lexical or morphosyntactic differences.This can be observed in the following examples.This behaviour seems to indicate that V1 and V2 establish morphosyntactic relations in independent domains.

Adjacency between the NA and V2
In Section 3.1, we indicated that there is a difference between V1 and V2 with respect to how they relate to the NA.Precisely, while there can be a 4-level break between V1 and the NA, the adjacency between the NA and V2 cannot be interrupted by such a break.

/
This asymetry between verbal duplicates has a syntactic counterpart.In this respect, as it is shown in (36), it is possible to add one constituent between V1 and the NA.However, (37) shows that the same constituent cannot interrupt the adjacency between the NA and V2.
( There is another similar behaviour.This is related to the combination of a non-local verbal doubling with the counterexpectation particle si, which is typical in PatSp (Pellejero and Silva Garcés 2015).From a semantic point of view, counterexpectation particle si eliminates a potential conversational implicature coming from the interlocutor.From a syntactic angle, a key feature of this discursive particle is its adjacent position to the XP NA , as shown in the following data.
(38) ¿Salimos a caminar?'Do you want to go for a walk?' Int.'I hope that they cook it whole!'

Semantic Import
The semantic characteristics related to non-local doublings in PatSp are somehow paradoxical.On the one hand, there seem to be no differences between the propositional meaning of a sequence without doubling and the one with duplicated elements (48).On the other hand, sequences with verbal doublings and sequences without them cannot occur in the same contexts.For instance, while constructions without duplicates can be the answer to a wh- In general terms, we could claim that verbal duplicates express the non-expectation and surprise that the proposition could cause on the listener, depending on the knowledge that the speaker assumes to share with his/her interlocutor.In this respect, (50a) can be paraphrased as in (50b).'There were guanacos and I understand that this could be unexpected for you' In this work, we claim that in the duplicated verbal constructions, there are two layers of meaning: one that is at-issue and another one that is not-at-issue (see Potts 2007).An at-issue meaning corresponds to the informative content of the sentence, that is to say, to the main information that the speaker wants to convey.At-issue content (henceforth, p) is equivalent to the proposition coded in the sequence, and it is not presupposed.In contrast, non-at-issue meaning (henceforth, m) conveys content that is secondary or irrelevant to the truth conditions of p. Accordingly, m indicates that p may be unexpected for the listener and, as a consequence, it can cause surprise. 8 That p and m constitute two independent layers of meaning can be evidenced by certain behaviours.Firstly, the listener can react distinctively either in relation to p or m.The different answers in B can be compared in the following example.
(51) A: Más allá hay otro pueblito, Treneta se llama, está CERQUITA está 'There is another little village, its name is Treneta, it's really near!' B1: Eso no es verdad, Treneta está muy lejos de acá → ¬ p 'That is not true, Treneta is really far away' B2: Está bien, pero eso no tiene nada de sorpresivo → ¬ m 'Ok, but that is not unexpected at all' Secondly, m cannot be refuted by the same speaker who produces the verbal doubling without resulting in a semantic contradiction.In (51), then, A cannot deny the surprise that the sequence could cause on the listener.
(52) Más allá hay otro pueblito, Treneta se llama, está CERQUITA está, #y no creo que te resulte sorpresivo 'There is another little village, its name is Treneta, it's really near! #And this is not unexpected for you' Thirdly, unlike p, the meaning of m cannot be retrieved in an ellipsis process, as is shown in the following contrast.
(53) a.The above-mentioned semantic traits indicate that m-i.e., non-expectation and surprise-is not part of the coded proposition in verbal duplicates.In this sense, it could be argued that m does not have an assertive semantic content but one that is suggested by the speaker.In other words, in verbal non-local doublings, it is implied that p is surprising, but it is not explicitly stated.
In order to conclude this section, it is worth mentioning that the surprise element is not attributed to the speaker, but to the listener.Particularly, the layer of meaning m is systematically built up thanks to the expectations that the speaker has in relation to his/her interlocutor.This can be demostrated by means of two other characteristics: verbal doublings in PatSp are not compatible with 2nd person singular or plural (54), or with imperative construcctions (55).( 54 The incompatibility between verbal doublings and the 2nd person can be accounted for in a direct way if the source of the surprise and the non-expectation is assumed to belong to the shared knowledge that the speaker assigns to the listener.

Summary
We now sum up the descriptions developed so far in terms of the prosodic, syntactic, and semantic aspects of verbal non-local doubling in PatSp.Table 1 offers a summary of the behaviours and characteristics found in our data.The evidence provided in the sections above constitutes the basics for the analysis that follows.

Analysis
In the generative literature, the phenomena related to non-local doubling have been explained mainly in terms of the copy theory (Chomsky 2000;Nunes 2004;Corver and Nunes 2007).The analyses conducted in this framework reveal that doubling is the materialisation of the same syntactic object that is placed, at the same time, in two different positions in the hierarchy.The reason for this double manifestation is structural: one syntactic object X escapes from TP/ΣP and moves up to the C domain; then, the remnant of TP/ΣP moves to a higher structural position.In this configuration, none of the copies of X are deleted (Nunes 2004;Muñoz Pérez 2017).The escape movement of one of the duplicates and the movement of SΣ/ST are illustrated in what follows with an example taken from RSp.As is shown in the diagrams in (56b), initially, V moves to C (by means of v, T, and Σ).Then, ΣP moves to the highest position in the hierarchy (Spec,CP).From this position, the copy of the verb in Spec,CP does not c-command the copy of the verb in C; this is why they cannot be deleted.This strategy (with minor changes) has been applied in the study of non-local doubling in different languages (see, for example, Jokilehto 2016 for Italian and Cvejanov and Druetta 2020 for Argentinian Sign Language).
Despite the explanatory potential of the copy theory, it cannot account for the data in PatSp.There are at least two empirical reasons.Firstly, proposals within the copy theory predict that the duplicated elements have to be identical, as exemplified in (56a).However, as we develop in Section 3.2.3,duplicates in PatSp can have different morphological features and roots.Secondly, from the analysis in (56b), it follows that no element can appear on the right margin of the second duplicate because the lowest ΣP in the hierarchy is deleted.In accordance with Muñoz Pérez (2018), this is the reason why instances such as (57) are ungrammatical in RSp.Unlike RSp, in PatSp, some elements can occur on the right of V2, as we show in Section 3.2.2(in fact, a case such as ( 57) is perfectly grammatical in PatSp).
These two empirical reasons and others that are of a conceptual nature 9 lead us to rule out the copy theory of movement as a possible explanation of non-local doubling in PatSp.The proposal that we put forward here adopts a radically different point of view.Precisely, we argue that duplicated sequences are not monoclausal but biclausal.In what follows, we present the key hypotheses of our analysis.
(58) First hypothesis Verbal non-local doublings in PatSp are built with two different clauses, CP1 and CP2, and V1 and V2 belong to independent syntactic domains.
(59) Second hypothesis In CP2, XP NA moves to the left periphery of the clause, i.e., it is focalised.
In Section 4.1, we provide findings from research studies that support our proposal.In Section 4.2, we explain how our two hypotheses can account for the behaviours in doublings in PatSp.

Background for the Hypotheses
The hypotheses underlying our analysis of doublings in PatSp are based on recent research on right dislocation (henceforth, RD) and focalisation.This subsection is aimed at showing previous studies in this respect.
As for the first hypothesis, it predicts that non-local doublings are biclausal constructions.Biclausality has been the focus of attention in order to explain recent findings on RD in Mandarin (Cheung 2015), Germanic languages (Truckenbrodt 2015;Ott and de Vries 2016), and Romance languages (Estigarribia 2020;Fernández-Sánchez 2020).One example of RD is shown in ( 60).In an RD, a constituent XPd that carries given information (Schwarzschild 1999) is placed on the right margin of a clause (CL) that is prosodically, semantically, and syntactically complete.In general terms, XPd has a correlate in CL, and RD has the following structure.
(61) [ CL ... correlate i ... ] XPd i The consulted literature highlights the fact that XPd is structurally independent from CL.In relation to prosody, XPd is a deaccented constituent that separates from CL by means of a special prosodic change.In (60), this change is indicated with a comma (hence the term comma intonation; see Potts 2007).From a semantic point of view, the presence of XPd is irrelevant to the truth conditions of CL.This means that CL (62a) and CL + XPd (62b) are semantically equivalent, as (62c) indicates.
(62) a. Tasman heeft ze gezien b.Tasman heeft ze gezien, die Maori's c. ∃x,∃y, x=Tasman, y=ze, x heeft gezien y (ze=die Maori's) Lastly, in syntactic terms, XPd does not play a role in the syntactic relations of CL.In order to illustrate this point, the example in (63a) can be considered, in which the coreference between ella and the XPd Ana does not violate Principle C, a behaviour that directly contrasts with the version in (63b), in which XPd is part of CL (as can be observed in its prosodic aspect).

Ana i Ana
However, right dislocations surprisingly behave as if XPd was part of CL.This has been referred to as connectivity effects (Merchant 2004).One of these effects has to do with case features in those languages in which this information is morphologically realised.The following example illustrates this.As shown in (64), dative case in XPd dem Peter (explicit in the article dem) has to match with dative case in the correlate ihm.This behaviour seems to suggest that case assignment for XP NA takes place in CL.
Another similar behaviour is related to reconstruction effects.This is shown in the following example.Ana's sweater 'She *i/j put it in the dryer, Ana i 's sweater' In (65), subscripts indicate that coreference between pro and Ana is ungrammatical.This behaviour seems to show that in the LF el suéter de Ana, CL is reconstructed in the position of its correlate, which would violate Principle C.This Janus-faced trait of RD can be explained if we assume that these constructions are formed with two semantically parallel clauses, CP1 and CP2, and that XPd is a fragment, i.e., the remnant of a clausal ellipsis process in CP2. 10 From this perspective, in (60), the underlying structure is the following.
(66) [ CP1 Tasman heeft ze gezien ] [ CP2 die Maori's i [ /////// heeft////////// Tasman/// t i ///////// gezien ] ] We have argued so far that, in accordance with the biclausal analysis of RD, non-local verbal doublings in PatSp are constituted by two clauses, CP1 and CP2, in which V1 and V2 correspond to different clauses.Additionally, we claim that CP1 and CP2 are semantically parallel.Now, there are at least two differences between RD and verbal doublings in PatSp.The first one is semantic.While in an RD, XPd constitutes a paraphrase, an explanation, or a richer description of the correlate, in verbal doublings in PatSp, the duplicated sequence is associated with mirativity, as developed in Section 3.3.The second difference is syntactic.Unlike what happens with RD, in verbal doublings in PatSp, a process of clausal ellipsis does not take place in CP2.In particular, we argue that V2 is not a fragment, as is the case for XPd.This is precisely our second hypothesis: in CP2, focus fronting occurs, i.e., movement of XP NA to the left periphery of the clause, which results in the mirative value that has been identified.Similar findings can be observed in studies by Frey (2010), Bianchi et al. (2016), and Cruschina (2019).These works attest that, in different languages, focus fronting results in an emphatic or mirative interpretation.The following examples can be considered.
(67) German (Frey 2010(Frey , p. 1424, his , his  In the data of ( 67)-( 69), the fronting of the XP NA (papayas, un anello di diamanti and a las Maldivas) brings about some alternatives to XP that are relevant to the interpretation of the whole sequence.This idea can be retrieved in ( 70 Considering ( 70), then, the fronting of XP NA allows to identify a set of alternatives (M) that are salient in a given discursive context.In M, the denotatum of XP is one the options available.The reviewed literature affirms that the A' movement of XP triggers a conventional implicature, according to which the proposition contained in the duplication is not the most expected one according to a stereotypical ordering of the alternatives in M, defined in line with the normal course of events (Kratzer 2012).Bianchi et al. (2016, p. 14) characterise the mirative implicature as follows.
(71) The proposition expressed by the clause is less likely than at least one distinct alternative proposition w.r.t. a contextually relevant modal base and stereotypical ordering source.
Taking into account ( 69), in what follows, we illustrate these proposals.When A las Maldivas fueron de luna de miel is uttered, the movement of a las Maldivas to the left periphery of the clause brings about some alternatives at a propositional level.These options share the same background, and they differ in the focalised element.73), the set of alternatives due to the fronting of a las Maldivas is p 1 , p 2 , p 3 , p 4 , p n , and the underlying proposition of the sequence of ( 69) is one of the options.Following Cruschina (2019, p. 140), the number of relevant alternatives depends on the context and the shared knowledge of the participants in a particular communicative situation.The key point in the proposal is that the mirative import projects a hierarchy of alternatives according to the expectations of the speakers and the normal course of events (Kratzer 2012).In this ranking, the underlying proposition is less probable than one of the options in the set.The contrast between the expectations of the interlocutors and the statement in ( 72) results in a mirative meaning.
In short, our analysis of verbal doublings in PatSp is based on two claims: they are biclausal sequences (such as RD) and, in the second clause, fronting of the constituent that bears the NA takes place (similarly to German, Italian, and Spanish, as studied by Frey 2010; Bianchi et al. 2016;and Cruschina 2019, respectively).We can assert, then, that verbal non-local doubling in PatSp presents the structure in ( 11), repeated below in ( 74).
We suggest that the (most of the) behaviours and traits described in Section 3 can be accounted for by the structure in (74).

Biclausality
Let us first consider the hypothesis related to the biclausal nature of the analysed doublings.Claiming that the underlying structure of non-local doubling is (74) implies that between V1 and V2 there are no derivational relations.In fact, we suggest that the duplicated verbs belong to different clausal domains and, as a consequence, V1 and V2 are not related in the syntactic derivation.
The derivational independency between V1 and V2 allows us to explain three of the morphosyntactic properties described in Section 3.2.V1 and V2: The possibility that V2 carries morphological and lexical information that is different from that of V1 is directly related to the non-at-issue semantic aspect associated with nonlocal doubling.In accordance with the proposals about RD developed above, CP2 encodes a secondary meaning that corresponds to the semantic background and that, as a consequence, is irrelevant to the truth conditions of CP1. 11In this respect, CP2 presents metacommunicative information directed to the participants in a particular situation (Schneider 2015, p. 288).In the case of verbal non-local doubling in PatSp, we argue that the morphological and lexical differences can be explained by means of the semantic nature of CP2 and its possibility of encoding secondary information (in this case, declarations and corrections).
In relation to the second point, a datum such as ( 26), repeated in (77a), would have the structure in (77b) according to our proposal.If (77a) actually has the structure in (77b), it is expected that V1 and V2 can establish subject-verb agreement relations in an independent way.The reason for this is that, in our view, CP1 and CP2 present different subjects and, even though features of person and number in both subjects are typically the same (see below), identical manifestation is not obligatory.Along these lines, the structure of (78) could be specified as follows. (78 Finally, the third point indicates another relevant behaviour in verbal doublings in PatSp.Since V1 and V2 are part of different clauses, it is expected that both predicates select their arguments independently.In (78), for example, we observe that while the external argument of V1 is 1SG, the external argument of V2 is 1PL.
As we saw in Section 3.2.3,duplicate verbs are equivalent in argument terms, that is, they select the same number of arguments and assign the same thematic roles.Now, if this is so, it is necessary to explain what happens in CP1.Consider again the structure of (78).Both V1 and V2 are dyadic predicates, so each of them selects two arguments in their clause domains.As shown (79), the argument structure of V1 is incomplete because one of its arguments does not materialise (a situation that we represent by '...' in the data).

b. V2 trajimos → agent = 1PL, theme = tres percas
In this paper, we argue that CP1 is an incomplete clause only in appearance.In (79), V1 selects a theme (in addition to an agent), which is semantically and syntactically equivalent to XP NA in CP2, as shown in (80).
(80) V1 traje → agent = 1SG, theme = tres percas One of the reasons for arguing that CP1 is an incomplete clause is the Projection Principle (Chomsky 1981(Chomsky , 1986)), according to which syntactic structure is a projection of lexical requirements, while the subcategorisation properties of lexical pieces must be satisfied.In this sense, if the argument requirements of V1 traje were not satisfied, (77a) would be an ungrammatical sequence, contrary to the facts.In Section 4.2.3, empirical evidence is offered for this statement.

A' Movement in CP2
Our second hypothesis is that XP NA moves A' on the left periphery of CP2.In other words, we assert that in CP2, a process of focalisation takes place.Most of the behaviours described in Section 3 can be directly accounted for with this claim.
In relation to prosody, one of the characteristics of focalisation is prominence or emphatic stress (Leonetti and Escandell Vidal 2021, p. 100).In verbal doubling in PatSp, this prosodic trait is manifested with the tonal configuration L+H* L-, in which there is a rising movement followed by an abrupt fall.
Focalisation of XP NA , moreover, implies the obligatory adjacency between the NA and V2.This characteristic is directly explained if we assume, together with Torrego (1984), Rizzi (1996), and related literature, that in the process of focalisation, V rises up to C, similarly to what happens in a whmovement.The tree diagram in (81), adapted from Kotzoglou (2006, p. 97), illustrates this movement.The result of this configuration presents the following order: XP NA -V-subject.The structure in (81) allows us to explain not only the adjacency between the NA and V2, but also the obligatory subject inversion and the impossibility of introducing linguistic material between the NA and V2.
In addition, with respect to the prosodic configuration of (81), XP NA and V2 cannot be separated by an interruption in phonation.This behaviour has been pointed out by Fábregas (2016) for cases of focalisation in general Spanish.As the minimal pair in (82) indicates, there cannot exist a prosodic break (as represented by comma in (82b)) between the focalised constituent and the verb.
(82) General Spanish (Fábregas 2016, p. 33, adapted  (84) General Spanish (Leonetti and Escandell Vidal 2021, pp. 60-61, their examples (39) The possibility of V2 subordination, nonetheless, seems to be semantically and syntactically conditioned by the properties of the subordinating verbs.In semantic terms, with verbs such as creo and parece in (85), the speaker mitigates his/her commitment on the propositional value of the sequence (expected values due to the not-at-issue nature of CP2; see Schneider 2015, p. 287 et seq.).Syntactically, subordinated sentences with verbs such as creer and parecer present main clause properties (Fábregas 2016, Section 6).For example, they allow for the use of speaker-oriented adverbs (86a) and the topicalisation of the verbal phrase (86b).( 86 The fact that the verbal duplications of PatSp cannot occur in subordination contexts is evidence in favour of the claim that these sequences involve a focalisation process. It has also been pointed out that focalisation is not compatible with any sentence modality (Leonetti and Escandell Vidal 2021, pp. 60-61).Sentences that have a focalised constituent cannot occur in interrogative or desiderative sentences, as the examples in (87) show.
(87) General Spanish Int.'I hope that we go until the WILLOW' As we have seen in Section 3.2, the same behaviours are observed in non-local doubling in PatSp.
In short, there is sufficient empirical evidence to support that in verbal doubling of the studied variety, the XP NA moves to the left periphery of CP2.Now, it is worth analysing how this movement triggers the mirative import associated with sequences that include a duplication.
As we discuss in Section 3.3, non-local doubling presents two layers of meaning: a propositional meaning, p, and a secondary meaning, m.Unlike p, the semantic value m is not an asserted content, but an implied one.This means that m is not obtained compositionally, i.e., it does not arise from the sum of the meanings of the elements that make up the sequence, rather, it is an implicit content systematically encoded in a specific syntactic structure (for an in-depth discussion on the characteristics of conventional implicatures, see Fernández Ruiz 2015, 2018).In verbal doubling of PatSp, the implicit meaning encoded in the duplication is that p could be unexpected for the listener.
In Section 4.1, we discuss the way in which focalisation encodes a mirative semantic value in Italian (Bianchi et al. 2016), German (Frey 2010), andPeninsular Spanish (Cruschina 2019).In this paper, we argue that this reasoning can be applied to the verbal doubling in PatSp.
Let us see an example.In (88), the values p and m can be differentiated, as (89) shows.In line with the works by Krifka (2007) and Krifka and Musan (2012), we understand that focus signals a set of relevant alternatives for the interpretation of the entire sequence.Indeed, in CP2, the focalisation of many people triggers the set of alternatives M described below.} members of the set M, that is, the alternative propositions of p, are ordered in line with a probability scale that operates in the listener's interpretation process, according to the state of knowledge that the speaker attributes to him/her.Thus, for the speaker, the state of knowledge of his/her interlocutor at the moment prior to enunciating the sequence containing the duplication is such that p1 is more likely than p2, p2 is more likely than p3, etc.The value m emerges precisely from the contrast between the order of the alternatives in M and the fact that the alternative that is finally stated is not the most probable option.
The hypothesis of focalisation in CP2, as we have seen, may account for the way in which the additional layer of meaning present in verbal non-local doubling in PatSp arises.This is certainly an advantage of our proposal compared to other analyses that exist in the literature on non-local doubling.
In addition, the hypothesis of focalisation in CP2 allows us to explain one of the fundamental properties of the analysed data: the condition of non-adjacency between the duplicates.In our analysis, the non-adjacency between V1 and V2 is a direct consequence of the movement of XP NA to the left periphery of CP2, which is required to trigger the characteristic mirative value of duplications.In other words, V1 and V2 are not contiguous because the XP NA systematically stands between them.Thus, the order V1-XP NA -V2 described in Section 3.2 and repeated here in (92a) can be reinterpreted as in (92b) in the light of our proposal.
If this analysis is correct, the non-adjacency between the duplicated verbs is not in itself a formal (morphosyntactic or prosodic) requirement for the good formation of verbal doubling in PatSp.Instead, it is a corollary of the way mirativity is encoded in this type of sequence.

On the Prosodic Materialisation of Doublings
To conclude this section, we briefly comment on some aspects related to the materialisation of doubling in PatSp.According to Match Theory, the syntactic structure determines the phonological representation in such a way that the units in Syntax have a direct prosodic correlate.Thus, while constituents are projected into phonological phrases (ϕ), (non-subordinate) clauses are projected into intonation phrases (ι).
(93) Match Theory (Selkirk 2011, p. 439) a. Match clause A clause in syntactic constituent structure must be matched by a corresponding prosodic constituent, call it ι, in phonological representation.
b. Match phrase A phrase in syntactic constituent structure must be matched by a corresponding prosodic constituent, call it ϕ, in phonological representation.
Therefore, Match Theory argues in favour of the following correlation: (94) One CP ⇐⇒ one IP If the correlation in ( 94) is correct, then our data exhibit a paradoxical behaviour.On the one hand, as observed in Section 3.1, verbal non-local doubling in PatSp is generally expressed in a single IP.On the other, these sequences are made up of two independent CPs, given the ample empirical evidence presented in previous sections.Then, our data seem to go against Match Theory and the one-to-one relation between CP and IP by establishing a reformulation of (94) in the following terms.
Two CPs ⇐⇒ one IP One way to resolve this (apparent) paradox is to propose that CP1 is an incomplete clause, because the element that receives the NA is not pronounced (something we express by means of '...' in (90)).
There is empirical evidence to support the claim that in CP1 there is silent material.The following example can be examined, in which the quantifiers ∀ and three interact.The sequence in ( 96) receives an ambiguous interpretation, as shown in ( 97).
(97) a. Interpretation 1: ∀ > three 'For all x, x = constructor, it is true that x built three houses' b.Interpretation 2: three > ∀ 'There is three houses, such that for all x, x = constructor, three houses were built by x' Where does this ambiguity come from?To answer this question, it is necessary to consider the two clauses that, in our analysis, form the doubling sequences.As can be seen in ( 98), CP2 only allows the second reading, according to which there are three houses that were built by all the constructors.The first interpretation, however, is not available in CP2.If interpretation ∀ > tres is not available in CP2, it follows that it must be available in CP1.This implies that CP1 contains quantifier tres in the domain of quantifier ∀.A plausible option is that this quantifier is part of the silent material in CP1, as shown in (99).In other words, it is possible to affirm that CP1 contains the quantifier tres that makes interpretation 1 possible.
The identity between the silent element in CP1 and XP NA in CP2 must also be syntactic.It can be demonstrated thanks to the categorical selection requirements of the predicates.To illustrate this observation, let us focus on the uses of the verb ganar(se) 'to place oneself', a typical verb in PatSp.
In a recent work, Garrido Sepúlveda et al. (n.d.) provide a detailed description of the grammaticalisation process of the verb ganar(se) in the Spanish spoken in Chilean and Argentine territory.
According to these authors, the verb ganar exhibits three well-defined grammaticalisation stages in Chilean Spanish.In the first one, ganar is used as a transitive verb with the meanings (perhaps the most widespread in the Spanish-speaking world) of 'to acquire some wealth or increase it', 'to earn a wage or a salary', or 'to obtain what is disputed in a match, game, contest, election, etc.'.In a second stage, ganar acquires a pronominal form, ganarse, and a locative meaning close to 'stay' or 'take place'.Finally, the most advanced of this process is the one in which ganarse loses (part of) its lexical content and it starts to denote an inchoative value as an auxiliary in an aspectual periphrasis.These values are systematised in Table 2.In short, there seem to be reasons that justify the assertion that a silent XP occurs on the right margin of CP1 with the same semantic values and the same syntactic nature as the XP NA of CP2.This is shown in (105). ( The fact that a silent element occurs in CP1 is the reason why verbal non-local doubling in PatSp materialises as a single IP.The reasoning is the following.According to Zubizarreta (1998), the NA is assigned by default through c-command relationships.Specifically, the NA is received by the constituent that is most embedded in the syntactic structure and, given Kayne's (1994) Axiom of Linear Correspondence, it is also the constituent that occurs in the right edge of the clause.This rule is known as the Nuclear Stress Rule.
(106) Nuclear Stress Rule (Zubizarreta 1998, p. 38) The lowest constituent in the asymmetric c-command ordering in the phrase is the most prominent in that phrase.
In this sense, the two CPs ⇐⇒ one IP correlation exhibiting non-local doubling in PatSp could be explained from the incompleteness of CP1.In other words, CP1 cannot be materialised as an independent IP because the most embedded constituent in CP1, i.e., the one that should receive the NA, is not phonetically realised. 12

Conclusions
In this work, we have put forward an analysis of verbal doubling in PatSp, a vernacular variety of Spanish spoken in Patagonia, Argentina.Our first objective has been to provide a prosodic, syntactic, and semantic description of the behaviours of this phenomenon.The accounts of the data have led us to identify a set of relevant features.In prosodic terms, a typical tonal configuration has been established: L+H* L-.This contour has not been registered in the inventory of prosodic patterns in RSp (Gabriel et al. 2010;Labastía 2018).From a semantic point of view, doublings in PatSp have a mirative interpretation.For the speaker, the underlying proposition of the sequence that contains the duplicated verbs could turn out to be surprising or non-expected for the listener.Lastly, in syntactic terms, we have discussed the interesting behaviours of V2 with respect to its structural relation with the previous sequence.V2 can present different morphological features or even different roots in comparison with V1 (which allows us to claim that duplicates belong to syntactically independent domains); simultaneously, the second duplicate seems to interact with the previous constituents in terms of argumental relations.In our view, most of the behaviours described in Section 3 had not been addressed in the literature regarding (non-local) doubling.
Our second objective has been to offer an analysis of the data and the particular behaviours observed.Our proposal consists of two central claims.In accordance with recent analyses of RD (Ott and de Vries 2016;Fernández-Sánchez 2020; inter alia), we argue that verbal non-local doubling in PatSp is not formed by one clause but by two clauses, CP1 and CP2.The second claim is that CP2 is the locus of the mirative value of duplications, which is syntactically encoded by the fronting of XP NA (an idea developed in Frey 2010, Bianchi et al. 2016, and Cruschina 2019).
Following these claims, we affirm that (105), repeated below, is the underlying structure in verbal non-local doubling in PatSp.We argue that the particular behaviours of this phenomenon in PatSp (or at least most of them) receive a comprehensive and more simple explanation from (107).
it is possible to formulate some lines of work for further research derived from the discussions in this article.One of them is related to the incompleteness of CP1.In Section 4.2.3, we argue that in CP1 there is a silent element that is semantically and syntactically identical to XP NA .Furthermore, we demonstrate that this property of CP1 is what guarantees that verbal doubling materialises in one single IP.However, the question remains open in relation to the nature of this silent material and the formalisation of the mechanisms that generate the incompleteness of CP1.
Along this article, for the sake of clarity, we have offered examples of verbal nonlocal doubling in the RSp, a Spanish variety that is very close to PatSp in geographic and historical terms (see Vidal de Battini 1964;Virkel 2004).Verbal non-local doubling in both varieties, as we have seen, share a number of properties (non-adjacency requirement, V1-XP NA -V2 order, emphatic/mirative value; see Muñoz Pérez 2017, 2018), but they differ in some other relevant behaviours (in PatSp, V1 and V2 may present morphological and lexical differences and different elements may occur on the right of V2, characteristics that are not present in RSp).Then, another line of research for a further study is to elaborate on a detailed comparison of the non-local doubling in both varieties of Spanish in order to determine whether (i) it is indeed the same phenomenon, (ii) the differences between both varieties can be explained in terms of microparametric variation (Kayne 2000), and (iii) the biclausal analysis developed here for PatSp can be applied to RSp data.
Dialectological research on PatSp has explained many of the morphosyntactic and phonological peculiarities of this variety in terms of language contact.Along this line, for example, Acuña andMenegotto (1993, 1995) explain the restructuring of number marking and pronominal systems of PatSp on the basis of the influence of the Mapuzungun on the Spanish varieties of the region.In the phonological domain, Virkel (2004, p. 203) understands that the assibilation of /r/ in PatSp is mainly observed in Spanish-speaking Mapuche people.Such considerations raise the question of the possible influence of Mapuzungun on the formation of the duplications studied in the previous pages.This enquiry will be addressed in the next stages of this research.
In relation to the prosodic analysis of the second duplicated element, our data indicate that there are no prominent syllables after NA in 94% of the analysed constructions.We have found only three instances of post-NA accents in the second ip, which could be related to two possible reasons not developed here.Post-NA accents could be due to (i) large phonological material between the accented post-NA word and its reference, or (ii) the lack of parallelism in the duplicated elements, as in habíamos mandado '(1PL) had sent' and mandamos '(1PL) sent', in which the speaker might feel the need to indicate that the two apparent different elements refer to the same event. 6 The term capicúa refers to a palindromic construction at the phrase level.7 See the Diccionario de partículas discursivas del español (http://www.dpde.es/#/entry/mejordichoaccessed on 16 August 2023), of Briz et al. (2008), online version.8 Bianchi et al. (2016, note 8) observe that non-expectation is also one of the values that is usually associated with exclamative sentences (see also Zanuttini and Portner 2003;Rett 2011;Torres Bustamante 2013).Thus, one might wonder if the PatSp doublings are not, in fact, exclamative sentences.The answer, however, is negative: the duplications studied in this work cannot be considered exclamations for two reasons.The first is that, unlike exclamations, in sequences that include a verbal doubling in PatSp, the propositional content is presupposed, not asserted.Hence, it can be denied independently.

Figure 1 .
Figure 1.Prosodic analysis of non-local doubling constructions in PatSp out of 68 cases.

Figure 2 .
Figure 2. Spectrogram of the sentence La Corona saben de ir mucha GENTE saben de ir 'Many people usually go to the Corona!'.

Figure 3 .
Figure 3. Spectrogram of the sentence Si es campo LINDO es 'It's nice land!'.
scientists a lot of animals many people the governor |−−focus−−| use to go (to the Corona) use to go (to the Corona) use to go (to the Corona) use to go (to the Corona) use to go (to the Corona) |− − −background− − −| (107) [ CP1 V1 [ XP ... ] [ CP2 XP NA i V2 [ TP t i ]] Se fueron por BARILOCHE se fueron los chicos]'The guys went by Bariloche!' B: Eso no es verdad, se fueron por Villa la Angostura 'That is not true, they went by Villa la Angostura' Some examples are shown below.
'The guys went by Bariloche!' 36) a. Se CULEBRAS snakes [ SP a to la the casa house ] llegaban arrive.IPFV.PST.3PL If the adjacency between the NA and V2 cannot be interrupted, it follows that verbal duplicates in PatSp and the counterexpectation si are not compatible together.This is evidenced by the impossibility of constructions such as (40).Incompatibility with Questions and Wish Clauses Non-local verbal doublings in PatSp are not possible in cases of questions and of clauses that express desire.This is evidenced in the ungrammatical constructions such as (46) and (47).
ENTERA entire la it.ACC.F cocinen cook.PRS.3PL question, non-local verbal doublings in PatSp are anomalous in the same context.
→ ∃x, ∃y, x=mi viejo, y=un campo, such that x has y & y is beautiful m → 'p could be unexpected for you' p → ∃x, ∃y, x=mi hermano, y=un campo, such that x has y & y is lindo m → 'p could be unexpected for you' b.Mi my p

Table 1 .
Synthesis of the prosodic, syntactic, and semantic behaviours of verbal non-local doubling in PatSp.
example (27)) A: Wat has Otto dieses Mal Besonderes auf dem Markt gekauft?'What extraordinary thing did Otto buy on the market this time?' B: PaPAyas i papayas ).As described in Section 3.2, the same behaviour can be identified in non-adjacent doubling in PatSp.However, the data presented in (44a) and (44b), repeated below, indicate that V2 can appear in cases of subordination.

Table 2 .
Meaning of ganar(se) in Chilean Spanish.Pronominal uses of ganarse with a locative value have been registered in PatSp.GarridoSepúlveda et al. (n.d.)point out that ganarse in this sense can only be combined with adverbial or prepositional phrases.The data in (102) are anomalous because, despite denoting location, the complement of ganarse is an AP, NP, or DP.With this in mind, let us resume our discussion on verbal doubling in PatSp.In the case of (103), in which the verb ganarse is duplicated, there is only one interpretation available: that of ganarse as a locative verb.This means that both V1 and V2 must be assembled with a PP/AdvP.The conclusion then is that, in CP1, '...' is a PP/AdvP.
*: We won X and we placed ourselves on this side; I understand that it sounds surprising to you.