A Formal Approach to Spanish ‘Genitive’ Pronouns in Non-Nominal Domains †

: This paper examines the distribution of ‘genitive’ pronouns in non-nominal domains in Spanish. These pronouns can alternate with constructions headed by the item de ‘of’ and a pronoun or other Determiner Phrases (DPs). In general Spanish, this alternation between a synthetic ( nuestro ‘our’) and an analytic ( de nosotros ‘of us’) option is found in the nominal domain. However, when looking at variation, this alternation appears in adverbial, verbal, and adjectival domains too. We discuss this phenomenon from a neo-constructionist approach, which assumes the late insertion of phonological exponents. We propose that the analytic and the synthetic options have almost the same syntactic structure, the only difference being the nature of the nominalizer’s ϕ -features. When the nominalizer values its ϕ -features, it can be lexicalized alone, and the ‘genitive’ pronoun lexicalizes the rest of the structure, including the introducer p /Place. Otherwise, when the nominalizer cannot (or needs not to) value its features, a ‘non-genitive’ pronoun lexicalizes the pronominal structure, and the head p /Place is lexicalized by the item de . Our proposal explains the complementary distribution between agreement/nominal morphology and the item de observed in many Spanish constructions. Different consequences are advantageously deduced.


Introduction
The so-called 'genitive' pronouns can be found in many grammatical contexts in Spanish when different varieties are taken into account.In most of them, forms like nuestro 'our' alternate with an analytical structure headed by the item de (de nosotros 'of us'), which also appears when there is a non-pronominal DP: de Juan/de la abuela 'of Juan/of the grandma'.This alternation has been studied extensively in the nominal domain but has received less attention in adverbial (1) and verbal domains (2) 1 .
'John remembered me' However, over the last years, many scholars have started to pay attention to these contexts, following pioneering research by Bertolotti (2014) and Mare (2014Mare ( , 2015)), who discussed them from a descriptive and a theoretical perspective, respectively.We can mention Bouzouita and Casanova (2017), Bertolotti (2017), Bouzouita (2020), Hoff (2020), Casanova and Pato (2022), Marttinen Larsson (2022), and Casanova (2023), among others.All these studies present a very detailed description of the different scopes in which the alternation is found.Moreover, they offer a wide range of examples in which it is possible to see that nuestro-and the rest of the paradigm-alternates with sequences other than de + pronoun, such as en 'in' + pronoun or sobre 'on' + pronoun.
The goal of this paper is to contribute to the study of this alternation from a theoretical perspective.Accordingly, we are not going to take it for granted that nuestro is a genitive pronoun or that de is a preposition.From now on, we will refer to both structures as the synthetic (nuestro) and the analytic (de nosotros) options because it is the most neutral way we have found to name them 2 .Whenever the traditional labels are required, quotation marks will be used.The neutrality in the terminology is motivated by the hypothesis that the variation observed is the result of options for lexicalization in which ϕ-features play an important role.
As we can infer from the data, there seems to be a kind of contention between the scopes allowing the synthetic option: [[[nominal domain] adverbial domain] verbal domain].The possibility of discussing them from Bobaljik's (2012) Contention Hypothesis is both attractive and challenging since 'nominal', 'adverbial', and 'verbal' are not primitives in themselves, as is well known.Understanding the pattern involved in the observed contention is my goal here.
The paper is organized as follows.After this brief introduction, Section 2 focuses on the description of the main data in order to define the characteristics that seem to be shared by the synthetic forms in the different scopes.Section 3 is concerned with previous approaches and the theoretical framework adopted in our research.In Section 4, the analysis of this phenomenon is developed in four steps: the structure of pronouns, the argument structures that trigger the alternation, the alternation per se and some issues on variation.Some final remarks are pointed out in Section 5.

The Data
Before delving into our data, we would like to present the general forms for the synthetic and the analytic options (SO and AO, respectively, henceforth), for readers not familiarized with Spanish.We are going to focus on postnominal SOs 3 , which, on the one hand, are the forms involved in the phenomena studied here, and on the other, are in complementary distribution with AOs in all Spanish varieties.As can be seen in Table 1, the SO suyo is a syncretic form that can refer to 3rd person (SG/PL) and to 2PL in many varieties.It also shows syncretism in the referent's gender as can be observed in the alternatives provided in the AO column.The SO is peculiar in different ways.First, it is not clear which word class it belongs to (see Picallo and Rigau 1999, among many others).Second, it is not easy to relate them to a unique thematic role in the nominal scope because they can refer to possessors, agents, themes, experiencers, and so on.This last characteristic is shared with the AO 4 .
'Juan sat down behind us' The other scope that allows SO∼AO is the verbal domain, in which the SO is found mainly in some Spanish varieties of Uruguay and Venezuela (see Bertolotti 2014;Mare 2014;Casanova 2020).
'Ana remembered you' Notwithstanding the valuable efforts to classify semantically the verbs involved in this alternation, it is remarkable-as noted by Mare (2014)-that there is a formal requirement: verbs (and adverbs) that allow the SO are the ones that introduce an argument denoting a human entity by the AO.There is nothing like a subgroup of them that allows the alternation and another one that does not.This is extremely important because the goal of defining semantic properties involves this kind of verb in general, regardless of the variation phenomenon discussed here.There are two different questions that need to be addressed, then.One of them should focus on the argument structure behind those verbs that present an argument unrelated to markers like accusative or dative.The other one should be why this kind of verb admits the SO when the argument introduced presents person features.The same reasoning applies to adverbs, and interestingly, in the adverbial scope, variation is found when the human argument is headed by an argument different from de 'of'.For instance, the adverb junto 'nearby' can present the SO (junto tuyo 'near you') only in the varieties that accept the AO (junto de Alex 'near to Alex'), not in those in which the complement of junto is headed by the preposition a (junto a Alex 'near to Alex').
This condition is relevant to revise the verbal domain because it has been noted (see Bertolotti 2017; Casanova 2020) that the SO can alternate with AOs not only headed by de, but also by sobre 'on ', en 'in' and a 'to'. Casanova (2023, p. 53) offers a ranking of possibilities, where de is followed by sobre, en, and a being the least frequently required items that allow the alternation SO∼AO.As we argue in Section 4, this can be explained under our analysis; however, further discussion of the variation in argument introducers is necessary.Nevertheless, as SO∼AO in the verbal domain is not a widespread phenomenon, it is a methodological mistake to describe the introducers involved in the AO just following general Spanish descriptions.It could be the case that, as in our example of the adverb junto 'near', there would be variation in the introducer, and only the varieties that admit de in the AO would admit the SO.
Leaving this methodological observation aside, we now turn to take some of the examples found in the literature.We use GS for general Spanish in the AO just to leave the above-mentioned possibility of variation open.In (6), we present examples of the SO∼AO where the introducer de is in the AO, and in ( 7), examples in which other introducers can be found.
'Ana talks about me' d.Laura se recostó mío/sobre mí (GE) 'Laura lay down on me' Note that the alternative introducers in the AO (7) are not related to their more specific locative meaning, and when they are (7d), it is the verb that codifies this locative meaning, as the English translation clearly shows: recostarse 'lie down'; acercarse 'come up to'; separarse 'break away from'.Moreover, the contrast between (6b) and (7b) shows that the verb pensar 'to think' admits two items, de and en, depending on the presence of the manner adverb (bien 'well'/mal 'badly'), at least in GS.The case of hablar is also interesting because both introducers are possible, de (6c) and sobre (7c), but sobre in this context does not show its literal locative meaning.Bertolotti (2017, pp. 342-44)-as well as Casanova (2023, p. 46)remarks that there would be a possible analogical extension from the SO∼AO with the introducer de to the SO∼AO with other introducers.However, as mentioned, there are many other factors involved, and there is no diachronic evidence of expansion from one group to the other.This is the reason why we prefer to pay attention to the semantic properties of the introducers involved in the SO∼AO because their occurrence could be epiphenomenal (see Section 4.2).

On Reanalysis and Late Insertion
Bertolotti (2017) explores the SO∼AO, taking into account two mechanisms of language change: reanalysis and analogy.For the first, she follows Langacker's definition, in which the hearer plays a key role, owing to the assumption that ambiguity can trigger a resegmentation-changes that affect phonological exponents-or a reformulation-changes that affect semantics-.As regards reanalysis, Langacker (1977, p. 58) characterizes it as the 'change in the structure of an expression or class of expressions that does not involve any immediate or intrinsic modification of its surface manifestation'.Bertolotti proposes that the fact that SOs are not necessarily interpreted as 'possessive' items in the nominal domain entails a reformulation in the semantics of these items: the only relevant features are person and number.According to the author, this reanalysis would be the source for the expansion of the SO from the nominal domain to the adverbial and verbal domains via what she calls constructional analogy.In brief, reanalysis in terms of reformulation would semantically equate the SO and the AO, and consequently, speakers would employ SO by constructional analogy in the same domains as an AO is found.
In turn, Mare (2014) questions whether 'possession' would have ever been a semantic feature in SO because even the Latin equivalent forms played different theta roles in the nominal domain.The author follows the hypothesis that the SO is a personal pronoun, which encodes a case feature: [GEN(ITIVE)].The presence of [GEN] is not obtained due to semantic properties but to the fact that this pronoun merges in a particular domain.In contrast with Bertolotti (2017), Mare analyses [GEN] as a formal feature for arguments that are introduced out of the verbal domain.In her approach, the SO and the AO materialize the same syntactic structure, and this is the reason why they can alternate when a personal pronoun is involved.
Within the Distributed Morphology (DM) framework (Halle and Marantz 1993), Mare's proposal assumes the late insertion of phonological exponents.This means that in her analysis, Syntax manipulates features and creates syntactic objects by Merge, but phonological exponents are added later in the derivation.DM suggests that the structure created by the Syntax can be modified before the insertion of phonological exponents.This kind of change takes place in a component called Morphology (see Figure 1) and depends, on the one hand, on specific language' properties and, on the other, on the inventory of Vocabulary Items.
Within the Distributed Morphology (DM) framework (Halle and Marantz 1993), Mare's proposal assumes the late insertion of phonological exponents.This means that in her analysis, Syntax manipulates features and creates syntactic objects by Merge, but phonological exponents are added later in the derivation.DM suggests that the structure created by the Syntax can be modified before the insertion of phonological exponents.This kind of change takes place in a component called Morphology (see Figure 1) and depends, on the one hand, on specific language' properties and, on the other, on the inventory of Vocabulary Items.The general idea of Mare's proposal is that the phonological exponents for the SO are inserted when K(ase) [GEN] fused post-syntactically with person/number features on D(eterminer).Otherwise, the item de materializes K [GEN], and person/number features are materialized by the corresponding pronoun (D), giving rise to the AO.Language variation regarding the SO∼AO would be explained by the possibility of applying the post-syntactic operation Fusion in different scopes.In general Spanish, the fusion of K[GEN] and a D with person features would take place only in the nominal domain (un amigo nuestro 'a friend of us').In other varieties, it would take place in the adverbial domain also (cerca nuestro 'next to us'), and in some others it would take place in the verbal domain as well (se acuerda nuestro '(S)he remembers us').
Both approaches, Bertolotti's and Mare's, shed light on the SO∼AO but fail to explain some of the characteristics described in Section 2. First, both assume a feature that is hard to find out of the nominal domain: possessive or [GEN], respectively.Bertolotti solves this problem by assuming that a reanalysis of the SO has taken place.In turn, Mare moves between the complementary distribution of [GEN] and agreement for the adverbial and the nominal scope and 'defectiveness' for the verbal scope.Second, in Bertolotti's proposal, the SO∼AO is hard to explain when the item involved in the AO is not de, but en, sobre, or a (see examples in (7)).Mare's assumption that the relevant feature is [GEN] does not solve this problem either.Finally, neither of them pays attention to the containment relationship between the three domains that admit the SO∼AO.Although Bertolotti highlights a constructional analogy, there is no explanation for why the analogy moves in the way it does and why it does not move in any variety-as far as we are concerned-from the nominal domain to the verbal one, without going through the adverbial domain.
In the following section, we will revisit the ideas developed in both approaches from a late insertion perspective that explores a neo-constructionist approach to grammar.Specifically, we will explore an explanation for the presence of the SO in different domains inspired by Bobaljik's (2012) Contention Hypothesis, also called *ABA restriction (see Caha 2016, among others), to explain the containment relationship mentioned above.The *ABA effect states that in a three-member paradigm ordered by markedness, the pattern in which the first and the third share a form to the exclusion of the middle member is not attested.The general idea of Mare's proposal is that the phonological exponents for the SO are inserted when K(ase) [GEN] fused post-syntactically with person/number features on D(eterminer).Otherwise, the item de materializes K [GEN], and person/number features are materialized by the corresponding pronoun (D), giving rise to the AO.Language variation regarding the SO∼AO would be explained by the possibility of applying the post-syntactic operation Fusion in different scopes.In general Spanish, the fusion of K [GEN] and a D with person features would take place only in the nominal domain (un amigo nuestro 'a friend of us').In other varieties, it would take place in the adverbial domain also (cerca nuestro 'next to us'), and in some others it would take place in the verbal domain as well (se acuerda nuestro '(S)he remembers us').
Both approaches, Bertolotti's and Mare's, shed light on the SO∼AO but fail to explain some of the characteristics described in Section 2. First, both assume a feature that is hard to find out of the nominal domain: possessive or [GEN], respectively.Bertolotti solves this problem by assuming that a reanalysis of the SO has taken place.In turn, Mare moves between the complementary distribution of [GEN] and agreement for the adverbial and the nominal scope and 'defectiveness' for the verbal scope.Second, in Bertolotti's proposal, the SO∼AO is hard to explain when the item involved in the AO is not de, but en, sobre, or a (see examples in (7)).Mare's assumption that the relevant feature is [GEN] does not solve this problem either.Finally, neither of them pays attention to the containment relationship between the three domains that admit the SO∼AO.Although Bertolotti highlights a constructional analogy, there is no explanation for why the analogy moves in the way it does and why it does not move in any variety-as far as we are concerned-from the nominal domain to the verbal one, without going through the adverbial domain.
In the following section, we will revisit the ideas developed in both approaches from a late insertion perspective that explores a neo-constructionist approach to grammar.Specifically, we will explore an explanation for the presence of the SO in different domains inspired by Bobaljik's (2012) Contention Hypothesis, also called *ABA restriction (see Caha 2016, among others), to explain the containment relationship mentioned above.The *ABA effect states that in a three-member paradigm ordered by markedness, the pattern in which the first and the third share a form to the exclusion of the middle member is not attested.

Understanding the SO∼AO
There are at least two aspects to consider before we delve into the SO∼AO.The first one involves pronoun structure and features.In Section 4.1, we focus on this aspect, presenting an approach to Spanish pronouns that pays considerable attention to the possibility they have to agree or not.The second aspect pertains to the argument structures in which the SO∼AO is found.In the verbal domain, although the DP introduced by the item de can be interpreted as any accusative DP in terms of theta-role, it seems to be out of the canonical Probe-Goal relation with the v-√ or C-T complexes (Chomsky 2008).Further, in the adverbial domain, the DP introduced by de is related to locative meanings like SOURCE, GOAL, or just PLACE.It is the meaning of the adverb, not the meaning of the item de, that determines each interpretation of the DP.Both observations suggest that we should explore the formal properties of the structures that require the AO in order to clarify the syntactic contexts that trigger de insertion.Beyond the SO∼AO in other domains, the question should be why the item de introduces DP arguments so differently interpreted in terms of their theta roles.We explore an answer to this question in Section 4.2.Once the AO is understood, we will be ready to discuss the SO∼AO as proposed in Section 4. As previously mentioned, we follow Picallo and Rigau's (1999) proposal that SOs are pronouns, as they have personal features.Of course, they codify something else, which is generally understood as a 'genitive case' feature.To start, we are going to classify pronouns just following their visible morphological behavior.Accordingly, leaving clitics aside, two groups of pronouns can be distinguished in Spanish: agreeing pronouns and non-agreeing pronouns.Agreeing pronouns are the so-called 'possessive/genitive' pronouns and can present double number and gender information because, beyond their person, number, and gender deictic features, they need to agree anaphorically.Representational nouns, such as picture, allow us to observe this kind of double number/gender marking.As shown in ( 8), the pronoun agrees with the feminine singular noun foto 'picture', but the participle sentados 'seatted' shows gender and number information that seems to be copied from the internal pronoun features (masculine and plural).The agreeing group has four or five members, depending on the Spanish variety.The varieties under consideration have four members because 2PL is syncretic with 3SG and 3PL (Table 2).These four members need to satisfy an agreement-of-nominal-features requirement.When there is not a noun in their scope that works as a Goal for agreement, they present defective nominal morphology in many varieties 7 .As expected, there are many other varieties in which the pronoun seems to copy features from somewhere else (see Felíu Arquiola and Pato 2020 for variation on nominal morphology in adverbs).We focus on defective morphology, but our proposal also applies to other options (see Section 4.4).The non-agreeing group of pronouns presents the same deictic features and does not show nominal agreement with external syntactic objects.Spanish, contrary to other Romance languages, has complex elements for the 1PL nosotros (and the 2PL vosotros in the varieties that present this lexical item).The word nosotros is formed by nos and the adjectivelike element otros.This adjective-like element needs to copy gender and number features from a scope in which the number is always plural, and gender has deictic reference, as can be observed in ( 9), where the 1PL clitic nos can be indexed with both 1PL defective and 1PL feminine (see Note 13).
(9) Juan nos saludo a nosotros /nosotras Juan CL.1PL greeted to 1PL.DEF 1PL.FEM 'John greeted us' Differently from agreeing nuestr-, the agreement process inside nosotros/nosotras does never give contrastive features as the example in (8) because the nominal features are copied from the information present in the projections related to nos.The same explanation applies to 2PL vosotros/vosotras 8 .
The non-agreeing group can have more or fewer members depending on variation.As previously mentioned, the varieties under study do not present the vocabulary item vosotros for 2PL but ustedes.For 1SG, there are two items, yo and mí, the second being the materialization of 1SG under the scope of a prepositional element (see Mare 2015 for discussion).The same happens with 2SG vos/tú and ti.1PL presents the gender variation just mentioned (nosotros/nosotras), and the third person shows gender and number morphology (Table 3).As pronouns have person features, they are referential expressions and, hence, behave like DPs.Building on Panagiotidis (2002), we propose that all these pronouns, agreeing and non-agreeing, present a structure with a nominal categorizer, projections for person information, and the projection of number (#).Unlike Panagiotidis and in line with Vanden Wyngaerd (2018), we do not assume a projection of D since the projections involved determine the referential character of the pronouns.This point does not in any way affect our proposal for SO∼AO.
Number information can be obtained deictically or by person features hypermarking (Halle 1997): when the combination of opposite features occurs, the feature on # must be necessary [+PL].As 3rd person is always interpreted as human, we propose that the node Hum(an) is in the structure of all three personal pronouns, Harris's (1991) gender cloning taking place in this node 9 .
Regarding person, I employ the combination of Participant in Speech Event (Part) and Author in Speech Event (Auth), following Halle's (1997, p. 129)   The difference between agreeing and non-agreeing pronouns is in the nominal catego- The difference between agreeing and non-agreeing pronouns is in the nominal catego- The difference between agreeing and non-agreeing pronouns is in the nominal categorizer requirements.Mitrović and Panagiotidis (2020) propose that adjectives are bicategorial elements in all languages, and that is the reason why they behave both as nouns and as verbs.The authors argue against adjectivizers while defending the proposal that the elements that are recognized as adjectives across languages present a verbalizer and a nominalizer in their structure.Typological studies show that adjectives are categorially ambivalent because they do not equally combine nominal and verbal characteristics.Simplifying Mitrović and Panagiotidis' analysis, the authors argue that the predominance of one categorizer's characteristics triggers the excorporation of the other categorizer.In Indo-European languages, it would be the verbalizer in the complement of n the one that excorporates ( 17).
(17) Indo-European adjectives These abstract structures are materialized by different Vocabulary Items.A Vocabulary Item (VI from now on) is understood as a relation between syntactic-semantic features/parts of structure and phonological exponents.When Vocabulary Insertion takes place, it is said that a specific part of the structure has been lexicalized.The scheme in ( 16) represents a VI in terms of the Distributed Morphology framework.

(16) /phonological exponent/↔ [syn-sem features]
The difference between agreeing and non-agreeing pronouns is in the nominal categorizer requirements.Mitrović and Panagiotidis (2020) propose that adjectives are bicategorial elements in all languages, and that is the reason why they behave both as nouns and as verbs.The authors argue against adjectivizers while defending the proposal that the elements that are recognized as adjectives across languages present a verbalizer and a nominalizer in their structure.Typological studies show that adjectives are categorially ambivalent because they do not equally combine nominal and verbal characteristics.Simplifying Mitrović and Panagiotidis' analysis, the authors argue that the predominance of one categorizer's characteristics triggers the excorporation of the other categorizer.In Indo-European languages, it would be the verbalizer in the complement of n the one that excorporates (17).Going back to pronouns, it is clear that the members of both groups behave in all relevant contexts as arguments-i.e., as kinds-not as sub-events, and that is the reason why we do not consider that there is a verbalizer in their structure.However, agreeing properties can follow from the features of the nominalizer.Our hypothesis is that there are two kinds of ns: one with unvalued ϕfeatures (n [uϕ] ) and one without ϕfeatures (n) at all.n [uϕ] remains in situ while n merges internally to Hum, forming a complex node n/Hum.In the nominal domain, n [uϕ] satisfies this requirement by copying number and person features from the DP in its scope.The result of this is that agreeing features are lexicalized independently of pronominal information.In the adverbial and the verbal domain, there is no DP which n [uϕ] can agree with, but nominal morphology is also recognized.As will be argued in Section 4.4, this hypothesis explains the SO∼AO and seems to fit well with the empirical evidence of agreeing possibilities in non-nominal domains.

Arguments Headed by de
We will now move on to discuss argument structure assuming a neo-constructionist framework, according to which there is nothing such as lexical items that select one or more arguments.Conversely, argument structure properties and the interpretation of arguments arise from their locus in the syntactic structure.Acedo-Matellán ( 2016 Goal project in the structure hierarchically.In ( 19), the item to would lexicalize both Place and Goal, while in (20) from lexicalizes Place, Goal, and Source.
(19) To the house (20) From the house Beyond these technical observations, there are some relevant descriptive points to consider.First, the theta roles for the arguments involved in the SO∼AO are related, one way or another, with PlaceP in both the verbal and the adverbial domains.Second, although there are some restrictions in the SO∼AO in the nominal domain (see Note 4), they do not affect the theta-role interpretation for complements of Place, i.e., Themes and Locations.Third, as we observed in Section 3, there can be items different from de that introduce the DP, but interestingly, they are VI related to projections proposed for goal, source, or place interpretation.Finally, in all the cases, we find a predicative relation in which the argument is marked in some way.For the adverbial and the nominal domain, we do not expect to find, for instance, accusative case markers because there is nothing such as a v-√ that works as a Probe in a specific context (see Chomsky 2001Chomsky , 2008)).However, in the verbal domain, we need to explain why this kind of internal argument is not part of the Probe-Goal relation with a v-√.Demonte (1991) focuses on the so-called prepositional verbs in Spanish and distinguishes two major groups 11 according to the following pattern of behavior: A. the possibility of dispensing with the prepositional phrase (PP); B. agentivity; C. the option of being taken as a complement in a causative construction; D. PP extraction from weak islands.Interestingly, as follows from the results, the verbs under analysis behave exactly as one of Demonte's types: the type that presents an agentive DP and in which the PP is not a syntactic argument.Goal project in the structure hierarchically.In ( 19), the item to would lexicalize both Place and Goal, while in (20) from lexicalizes Place, Goal, and Source.
(19) To the house (20) From the house Beyond these technical observations, there are some relevant descriptive points to consider.First, the theta roles for the arguments involved in the SO∼AO are related, one way or another, with PlaceP in both the verbal and the adverbial domains.Second, although there are some restrictions in the SO∼AO in the nominal domain (see Note 4), they do not affect the theta-role interpretation for complements of Place, i.e., Themes and Locations.Third, as we observed in Section 3, there can be items different from de that introduce the DP, but interestingly, they are VI related to projections proposed for goal, source, or place interpretation.Finally, in all the cases, we find a predicative relation in which the argument is marked in some way.For the adverbial and the nominal domain, we do not expect to find, for instance, accusative case markers because there is nothing such as a v-√ that works as a Probe in a specific context (see Chomsky 2001Chomsky , 2008)).However, in the verbal domain, we need to explain why this kind of internal argument is not part of the Probe-Goal relation with a v-√.Demonte (1991) focuses on the so-called prepositional verbs in Spanish and distinguishes two major groups 11 according to the following pattern of behavior: A. the possibility of dispensing with the prepositional phrase (PP); B. agentivity; C. the option of being taken as a complement in a causative construction; D. PP extraction from weak islands.Interestingly, as follows from the results, the verbs under analysis behave exactly as one of Demonte's types: the type that presents an agentive DP and in which the PP is not a syntactic argument.
[PP omission] Beyond these technical observations, there are some relevant descriptive points to consider.First, the theta roles for the arguments involved in the SO∼AO are related, one way or another, with PlaceP in both the verbal and the adverbial domains.Second, although there are some restrictions in the SO∼AO in the nominal domain (see Note 4), they do not affect the theta-role interpretation for complements of Place, i.e., Themes and Locations.Third, as we observed in Section 3, there can be items different from de that introduce the DP, but interestingly, they are VI related to projections proposed for goal, source, or place interpretation.Finally, in all the cases, we find a predicative relation in which the argument is marked in some way.For the adverbial and the nominal domain, we do not expect to find, for instance, accusative case markers because there is nothing such as a v-√ that works as a Probe in a specific context (see Chomsky 2001Chomsky , 2008)).However, in the verbal domain, we need to explain why this kind of internal argument is not part of the Probe-Goal relation with a v-√ .Demonte (1991) focuses on the so-called prepositional verbs in Spanish and distinguishes two major groups 11 according to the following pattern of behavior: A. the possibility of dispensing with the prepositional phrase (PP); B. agentivity; C. the option of being taken as a complement in a causative construction; D. PP extraction from weak islands.Interestingly, as follows from the results, the verbs under analysis behave exactly as one of Demonte's types: the type that presents an agentive DP and in which the PP is not a syntactic argument.( 21 Gallego (2022) revisits Demonte's proposal and concludes that this type of prepositional verbs are unergative predicates (not special transitive predicates as in Demonte), in which the PPs are lexical adjuncts in Mateu (2002) terms: 'adjuncts related to the specific roots as part of their non-compositional (read 'encyclopedic') meaning' (Gallego 2022, p. 303).In general, we agree with Gallego's analysis, but we derive the idea of 'lexical adjuncts' from the syntactic structure so as to capture Demonte's intuition about transitivity.In an attempt to explain both properties, we focus on defectiveness as the impossibility of a head to project a Specifier position for an argument.The result of this is that the corresponding head merges in the structure, contributing to the relevant interpretation, but as a defective head, the argument that this head should introduce can be omitted, and if present, it must be introduced by another head in the scope of the main projection.The proposal on defectiveness has been explored in relation to Voice, the head that introduces DPs interpreted as agents (25a).When Voice is defective, the predicate does not lose the agentive interpretation, as, for instance, in Spanish periphrastic passive constructions (25b) or se constructions (see Pujalte and Saab 2012, for instance).Rather, the predication is interpreted as being caused by an agent that is not present in the main structure but can be reintroduced by a preposition (25c).
'The exams have been reviewed by Juan.' Our proposal is that in (25c), the PP por Juan 'by Juan' is an adjunct of Voice: the DP Juan is introduced in the structure by a p(redicative relator) merged in the domain of VoiceP.This is the reason why the DP is interpreted as an agent, and the item por 'by' cannot alternate with other prepositions: por is the Vocabulary Item that lexicalizes a p under the scope of VoiceP.Building on the idea of defectiveness, we would like to argue that most of the verbs under discussion, if not all of them, are the result of the impossibility of the head PLACE to introduce a DP argument.This means that the interpretation of the event is obtained compositionally thanks to the presence of that head (and the others involved), but the syntactic conditions for transitivity are not met because no DP merges in the relevant positions owing that PLACE is defective.Consequently, there is a required argument that can only be introduced in the area of this head-in fact, it will be interpreted as Theme or more specific kinds of locations as mentioned above-but it is hidden from the main structure's Probe-Goal relationships (Chomsky 2001).Accordingly, the unergativity proposed by Gallego is derived from a defective transitive structure, as in Demonte's analysis.We use p in (26) for 'predicative relation' in order to distinguish it from Place in the main structure, but both represent the same kind of relation.One of these verbs-although most probably not the only one-behaves like the other group of prepositional verbs in Demonte's classification: gustar (see Bouzouita and Pato 2019 for the SO∼AO with this verb).Since Demonte and Gallego agree that this group consists of unaccusative predicates that select a small clause with a PP-our PlaceP-(see Gallego 2022, p. 305), the conditions for the SO∼AO would be given without further ado.
Finally, the cases in which de seems to alternate with other items can be explained under the same system.In ( 26), for instance, if p merges with a root-like SOBRE or EN (see for instance (18b)), it acquires a more specific encyclopedic kind of relation; otherwise, the less specified item de lexicalizes p when no root is merged.We follow the same reasoning proposed for v lexicalization: some verbal elements like be, go and do just lexicalize a v in a particular predicative structure when no more specific roots are merged-or from other perspectives [BE], [GO] and [DO] are features or 'flavours' on this little v (see Folli and Harley 2005)-as in the contrasts between do a dance and dance.Furthermore, the SO would not alternate with an introducer different from de because these introducers are not primitives in the structure.In other words, the SO requires that there be a p/Place in the structure, not that this p/Place lexicalizes in one way or another.

The SO∼AO
Now that we have the basic ingredients let us move on to the SO∼AO.As discussed in Section 4.1, the arguments involved in the alternation are pronouns that can be materialized by different Vocabulary Items, depending on the syntactic structure and the features in it.On the other hand, the scopes in which the SO∼AO is found have in common that the DPs/pronouns this paper focuses on are introduced by the head Place/p that establishes a predicative relation.Out of the nominal domain, all these arguments are interpreted in a very restrictive way: they are either themes or locations.Last but not least, the general panorama shows that agreeing syntactic objects (examples in a) and One of these verbs-although most probably not the only one-behaves like the other group of prepositional verbs in Demonte's classification: gustar (see Bouzouita and Pato 2019 for the SO∼AO with this verb).Since Demonte and Gallego agree that this group consists of unaccusative predicates that select a small clause with a PP-our PlaceP-(see Gallego 2022, p. 305), the conditions for the SO∼AO would be given without further ado.
Finally, the cases in which de seems to alternate with other items can be explained under the same system.In ( 26), for instance, if p merges with a root-like SOBRE or EN (see for instance (18b)), it acquires a more specific encyclopedic kind of relation; otherwise, the less specified item de lexicalizes p when no root is merged.We follow the same reasoning proposed for v lexicalization: some verbal elements like be, go and do just lexicalize a v in a particular predicative structure when no more specific roots are merged-or from other perspectives [BE], [GO] and [DO] are features or 'flavours' on this little v (see Folli and Harley 2005)-as in the contrasts between do a dance and dance.Furthermore, the SO would not alternate with an introducer different from de because these introducers are not primitives in the structure.In other words, the SO requires that there be a p/Place in the structure, not that this p/Place lexicalizes in one way or another.

The SO∼AO
Now that we have the basic ingredients let us move on to the SO∼AO.As discussed in Section 4.1, the arguments involved in the alternation are pronouns that can be materialized by different Vocabulary Items, depending on the syntactic structure and the features in it.On the other hand, the scopes in which the SO∼AO is found have in common that the DPs/pronouns this paper focuses on are introduced by the head Place/p that establishes a predicative relation.Out of the nominal domain, all these arguments are interpreted in a very restrictive way: they are either themes or locations.Last but not least, the general panorama shows that agreeing syntactic objects (examples in a) and  27a), the adjective ballenera 'of whales' agrees in gender and number with la pesca 'fishing'; in (28a), the quantifier muchas 'many' agrees in gender and number with personas 'people'; and in (29a) nuestra 'our' agrees with una amiga 'a friend'.When each syntactic object presents its own gender and number features, the relation between them seems to be materialized by the item de.The same happens in the SO∼AO, with the difference that-at least in the varieties under study-it is not easy to see agreement markers because we find the default (DEF) nominal item -o.
( All things considered, it appears that SOs are not just person features, contrary to Bertolotti's (2017) proposal, but they represent phonological exponents that materialize not only the relevant person/number information but also the predicative projection that introduces it as an argument.In other words, the difference between nosotros and nuestro is that the first only materializes the pronominal features (33), while the second, more specifically nuestr-, lexicalizes pronominal information as well as the predicative relator (p/Place) (34).In fact, the proposal here is that the syntactic objects named adverbs or adjectives have the same property: they also lexicalize a relational head (or more than one).This fits well with Mitrović and Panagiotidis' analysis of adjectives because the nominalizer n [uϕ] relates to nominal features (gender and number), which are lexicalized independently by the corresponding vocabulary items.The lack of features on n motivates head movement to Hum, where deictic gender features are located 14 .The result is that when vocabulary insertion takes place, there will be just one node-the complex node n/Hum-for gender information lexicalization.Nonagreeing pronouns like nosotros never lexicalize PlaceP, and that is why this head/phrase must be lexicalized by other vocabulary items, de 'of' being the default one in Spanish.The same happens with other DPs, i.e., lexical arguments (de la mesa 'of the table') and proper nouns (de Juan 'of Juan).On the other hand, n[uφ] needs to value its features from its local domain, but whichever these features are, the information related to this n[uφ] lexicalizes independently.The second difference regarding lexicalization is that what we call agreeing pronouns also lexicalize Place (34), and this is the reason why they are not found as complements of prepositions or as sentence subjects or direct objects (consider the structures in Section 4.2).
In brief, the general requirement for triggering the SO∼AO is having structures like (33) and ( 34), where the only difference is the kind of n involved.If n has no features to value, it remains subsumed into the deictic information in the structure and is lexicalized together with the other projection.The vocabulary items that lexicalize this information cannot lexicalize Place, i.e., Place is not part of the structure and the features related to these phonological exponents.In contrast, when n is n[uφ], it must (somehow) value the relevant features, and once valued, n can be lexicalized.Hence, the rest of the structure is going to be lexicalized by a Vocabulary Item that also includes Place.This analysis captures Mare's (2014) observation about the complementary distribution between the presence of the item de and nominal markers.Furthermore, leaving aside the idea of possession or genitive case, it allows us to pay attention to a clear pattern related to the introduction of arguments: the SO∼AO occurs when a pronominal argument is introduced as a complement of p/Place.Our next goal is to explain the two kinds of variation observed: (1) restrictions on the SO∼AO with adverbs, adjectives, and verbs and (2) variation regarding nominal markers resulting from agreement.

On Variation and Beyond
Now that we have the bases for SO∼AO, it is time to discuss why SO occurrence is restricted in many varieties (35).The lack of features on n motivates head movement to Hum, where deictic gender features are located 14 .The result is that when vocabulary insertion takes place, there will be just one node-the complex node n/Hum-for gender information lexicalization.Nonagreeing pronouns like nosotros never lexicalize PlaceP, and that is why this head/phrase must be lexicalized by other vocabulary items, de 'of' being the default one in Spanish.The same happens with other DPs, i.e., lexical arguments (de la mesa 'of the table') and proper nouns (de Juan 'of Juan).On the other hand, n[uφ] needs to value its features from its local domain, but whichever these features are, the information related to this n[uφ] lexicalizes independently.The second difference regarding lexicalization is that what we call agreeing pronouns also lexicalize Place (34), and this is the reason why they are not found as complements of prepositions or as sentence subjects or direct objects (consider the structures in Section 4.2).
In brief, the general requirement for triggering the SO∼AO is having structures like (33) and ( 34), where the only difference is the kind of n involved.If n has no features to value, it remains subsumed into the deictic information in the structure and is lexicalized together with the other projection.The vocabulary items that lexicalize this information cannot lexicalize Place, i.e., Place is not part of the structure and the features related to these phonological exponents.In contrast, when n is n[uφ], it must (somehow) value the relevant features, and once valued, n can be lexicalized.Hence, the rest of the structure is going to be lexicalized by a Vocabulary Item that also includes Place.This analysis captures Mare's (2014) observation about the complementary distribution between the presence of the item de and nominal markers.Furthermore, leaving aside the idea of possession or genitive case, it allows us to pay attention to a clear pattern related to the introduction of arguments: the SO∼AO occurs when a pronominal argument is introduced as a complement of p/Place.Our next goal is to explain the two kinds of variation observed: (1) restrictions on the SO∼AO with adverbs, adjectives, and verbs and (2) variation regarding nominal markers resulting from agreement.

On Variation and Beyond
Now that we have the bases for SO∼AO, it is time to discuss why SO occurrence is restricted in many varieties (35).The lack of features on n motivates head movement to Hum, where deictic gender features are located 14 .The result is that when vocabulary insertion takes place, there will be just one node-the complex node n/Hum-for gender information lexicalization.Nonagreeing pronouns like nosotros never lexicalize PlaceP, and that is why this head/phrase must be lexicalized by other vocabulary items, de 'of' being the default one in Spanish.The same happens with other DPs, i.e., lexical arguments (de la mesa 'of the table') and proper nouns (de Juan 'of Juan).On the other hand, n [uϕ] needs to value its features from its local domain, but whichever these features are, the information related to this n [uϕ] lexicalizes independently.The second difference regarding lexicalization is that what we call agreeing pronouns also lexicalize Place (34), and this is the reason why they are not found as complements of prepositions or as sentence subjects or direct objects (consider the structures in Section 4.2).
In brief, the general requirement for triggering the SO∼AO is having structures like (33) and ( 34), where the only difference is the kind of n involved.If n has no features to value, it remains subsumed into the deictic information in the structure and is lexicalized together with the other projection.The vocabulary items that lexicalize this information cannot lexicalize Place, i.e., Place is not part of the structure and the features related to these phonological exponents.In contrast, when n is n [uϕ] , it must (somehow) value the relevant features, and once valued, n can be lexicalized.Hence, the rest of the structure is going to be lexicalized by a Vocabulary Item that also includes Place.This analysis captures Mare's (2014) observation about the complementary distribution between the presence of the item de and nominal markers.Furthermore, leaving aside the idea of possession or genitive case, it allows us to pay attention to a clear pattern related to the introduction of arguments: the SO∼AO occurs when a pronominal argument is introduced as a complement of p/Place.Our next goal is to explain the two kinds of variation observed: (1) restrictions on the SO∼AO with adverbs, adjectives, and verbs and (2) variation regarding nominal markers resulting from agreement.

On Variation and Beyond
Now that we have the bases for SO∼AO, it is time to discuss why SO occurrence is restricted in many varieties (35).
(35) a. Una amiga tuya (ok for all Spanish varieties) A friend of yours b. *Cerca tuyo (* for some Spanish varieties) Near of you c. *Me acordé tuyo (* for most of Spanish varieties) I remembered you d. *Estoy orgullosa tuyo (* for most of Spanish varieties) 15  'I am proud of you' Regardless of the variation observed, it is remarkable that there seems to be a kind of containment relationship among the domains that admit the SO.The varieties that present SO in the adverbial domain also show it in the nominal domain, and the varieties in which the SO is found in the verbal (and in the adjectival) domain are also found in the adverbial domain.

(36) [[[nominal domain] adverbial domain] verbal domain]
We have argued that the introduction of arguments that can be materialized by the SO∼AO in the three domains share the characteristic of being the complement of a predicative relator (p/Place).This is the first formal property that could motivate what Bertolotti (2017) calls constructional analogy.However, while the presence of this predicative relator is necessary, it is not enough to explain the containment represented in (36).
The analysis proposed by Mitrović and Panagiotidis (2020) shed light on this puzzle.As was remarked in Section 4.1, they argue against adjectivizers-or adverbializers-and propose that adjectives-as well as adverbs-are bicategorial elements with an n and a v in their structure.It means that, following the *ABA-effect way of reasoning, it is desirable, in fact, that a domain that presents a nominal categorizer triggers the SO∼AO before a domain that does not have n.Moreover, as the adverbial domain is also characterized by the presence of a verbalizer, the extension of the SO∼AO to an exclusive verbal domain also follows.Found and unfound (*) SO containment relations are shown in background color in Table 4.The challenge now is to define the features that motivate the different possibilities of SO insertion in Spanish varieties.It seems to be the case that SO insertion is conditioned by the possibility of n [uϕ] to value the relevant features.This kind of agreement operation takes place canonically in the nominal domain, where there is a nominal element with relevant features to value.In some varieties, when a structure like (34) merges in the scope of v, n [uϕ] cannot value its ϕ-features and, consequently, it is computed as an n, giving rise to the insertion of a non-agreeing pronoun.As already explained, the result of this insertion is the AO.However, let us consider some ideas that can help us to further understand this topic.
First of all, contrary to previous research, we argue that the presence of SOs in nonnominal domains is not due to the characteristics of SOs but to the 'solutions' provided by each variety when the nominalizer in a pronominal structure is a n [uϕ] .We propose that there are different solutions: Hypothesis 1.When n [uϕ] cannot value its features, the derivation is ruled out.It means that only the structure in (33) results in a good derivation.
Hypothesis 2. When n [uϕ] cannot value its features, n [uϕ] is computing as an n.It means that in Syntax, both (33) and (34) result in a good derivation.However, as n [uϕ] is computed as an n, conditions for insertion change, and SOs are ruled out for vocabulary insertion.
If we follow hypothesis 2, we can go a step further because if n [uϕ] is not ruled out, some varieties could admit a n [uϕ] in the structure until it can value features with other n, at least, a defective one.This could be the case of SOs in the adverbial domain.As pointed out by different authors (Bartra andSuñer 1992, 1997), some adverbs' morphology kicks in as a default option.Interestingly, it could also explain why the SO is not found with the same extension across varieties in the adjectival domain.As already mentioned, adverbs and adjectives are bicategorial, but SO is more frequent with adverbs than with adjectives.This could be caused by n features again: while adverbs present default ϕ-features on n-let us use [-F] and [-PL] for default-adjectives must value their ϕ-features inside the DP they merge with.This 'delay' could lead to the n [uϕ] of the pronoun being computed as an n, and we know what the result is: the AO is the only possibility for materialization 16 .
Finally, there would be varieties in which a n [uϕ] that cannot value its ϕ-features with a nominal element attributes them a defective value, again [-F] and [-PL].When this happens, it can trigger vocabulary insertion, and an SO can materialize the rest of the structure.
Table 5 summarizes the above-mentioned possibilities.We call V(ariety)1 the most restrictive variety, which only admits SOs in the nominal domain.V4 is on the opposite side: it admits SOs in the four domains.V2 admits SO with adverbs as well as in nouns, and V3 admits SOs with verbs also.

Adverbs
Moreover, hypothesis B sheds light on the variation observed in the examples in (37), which represents the ways in which n [uϕ] values its ϕ-features across Spanish varieties (see Kany [1945Kany [ ] 1970, p. 66;, p. 66;RAE-ASALE 2009, p. 1361 ), and the phonological exponent for these default features is /o/.The default option and the defective features can be different across varieties, and this seems to be the case in (37b), where the exponent /a/ lexicalizes the nominalizer.The third option is the deictic solution: n [uϕ] values its features with the deictic features in the pronoun (37c).Interestingly, none of these ways of solving the problem of having a n [uϕ] out of the nominal domain affects the interpretation provided by the Syntax.It only affects the conditions for vocabulary insertion.

Final Remarks
This paper proposes a theoretical approach to the SO∼AO in different domains in Spanish varieties from a neo-constructionist perspective that assumes the late insertion of phonological exponents.The analysis of these different domains allows us to find a pattern: when a pronoun merges as the complement of a predicative relator (p/Place) remaining out of the domain of a v or a T Probe-Goal relation, the AO alternative takes place, and consequently the SO∼AO can be found.Nominal, adjectival, and adverbial domains do not present the proper v and T for the Probe-Goal mechanism.For the verbal domain, following previous research on 'prepositional complements', we propose that most of the verbs involved in the SO∼AO are unergative verbs with a defective Place.Consequently, there is an argument-the one involved in the SO∼AO-that is not part of the main structure but introduced by a predicative relator in the scope of Place.This is the reason why, in the verbal domain, these arguments have a very restrictive interpretation in terms of thematic roles.
The other part of our argumentation concerns pronouns.In brief, the distinction between agreeing and non-agreeing pronouns was relevant to leave aside the classical features attributed to the SO, and consequently, we could explain the complementary distribution between nominal agreement and the presence of the item de.The conclusion was that de codifies p/Place when the p/Place complement does not have an n [uϕ] .Otherwise, as the n [uϕ] lexicalizes independently when it can value its ϕ-features, the alternative for SOs insertion is open in order to lexicalize p/Place.The way in which this lexicalization process works can follow from Fábregas' (2007b) Exhaustive Lexicalization Principle: all the nodes in a Syntactic structure must be lexicalized.
Last but not least, we advanced a hypothesis to explain the way in which the SO∼AO is expanded into different domains.We observed that there is some kind of containment between domains that reminds us of Bobaljik's Contention Hypothesis and the *ABA pattern.Following this way of reasoning, we proposed arguments against the idea that the SOs items have different properties from one variety to another, as in Bertolotti's proposal in terms of reanalysis, and alternatively argued that it has a close tie to the way in which every particular grammar solves the 'problem' of having an n [uϕ] in a domain in which it cannot value its ϕ-features.The observed contention between domains follows from that: in the nominal domain, there is an n with features to value; in the adverbial domain, there is an n with defective features; in the verbal domain, there is no n, but defectiveness is an alternative for this n [uϕ] .The adjectival domain is the most complex one, and as far as we are concerned, it seems to be the most restrictive for SOs.Our hypothesis is that this is because adjectives have to value ϕ-features in turn.
In sum, even though this linguistic puzzle has multiple parts, some patterns are very clear.A late-insertion approach has proven to cast light on some of these patterns, which, otherwise, remain lost in a mass of apparent exceptions.4 Although it is not the focus of this paper, it must be said that the distribution between SO and AO in the nominal domain shows some clear restrictions depending on the projections the pronoun merges into and the pronominal features involved: 1SG and 2SG are not possible options in the examples in (3a), (3b) and (3d), at least in most of Spanish varieties, while 3SG and person plural forms do not exhibit any restrictions.See Mare (2015, chp. 3) for an exhaustive discussion on this topic and also Fábregas (2014).

5
Moreover, when we focus on language variation, we also observe that the SO displays interesting differences on agreement morphology.See Marttinen Larsson and Bouzouita (2022) and references therein.In this example there is a transitive construction which also requires an argument introduced by de.The theta role of this argument is SOURCE, so this predicate presents an AGENT agreeing with the verb (subject), a THEME related to the accusative clitic la, and a SOURCE which trigger the SO∼AO in some varieties.7 This is not a special property of these pronouns, but a general property of adjective-like elements as can be seen in the contrast between altas 'high' (i) and alto 'high' (ii).Suñer and Di Tullio (2014) refer to cases like 'alto' in (ii) as bare adjectives and propose that their lack of agreement is due to the fact that they cannot match their features with a nominal element in their scope.We go back to this discussion in Section 4. The segment -tr-in nuestro and vuestro has no relation with the form otros 'others', but is the result of phonetic changes in Latin 'possessives' due to the omission of the [e] from nominative masculine singular noster (1PL)/voster (2PL) to other genders and cases: nostra, vostra, nostrum, vostrum and so on.9 In our system, the projection Hum(an) triggers the presence of person features such as participant and author and is equivalent to the feature [HUM(an)] in Harris (1991).The author proposes that [HUM] gives rise to a human cloning rule in Spanish that results in the introduction of the semantic sex features [MALE] and [FEMALE].These are the features that later in the derivation motivate the introduction of grammatical gender features [+FEM]/[-FEM].10 Although many authors defend person features' containment, it is possible to observe across languages that person syncretism works differently when number is plural or singular.For instance, when # is [-PL] syncretism between 1st and 3rd person is found in Mapuzugun 'possessive' pronouns and in Spanish subjunctive verbal tenses.However, when # is [+PL] the syncretism is found between 2nd and 3rd in most of Spanish varieties (see Mare 2021).For reasons of space, we will not delve into this topic in the present paper.According to Demonte (1991, pp. 72-73), the two major groups are (i) and (ii): i. consistir en 'consist on', prorrumpir en 'burst into', constar de 'consist of', abundar en 'abound in', adolecer de 'suffer from', versar sobre 'deal with', alardear de 'boast about', redundar en 'be redundant', reposar en 'lie', etc. ii.insistir en 'insist on', discrepar de 'disagree with', incurrir en 'commit', abusar de 'abuse of', prescindir de 'dispense with', alardear de 'boast about', pugnar por 'fight for', renunciar a 'resign ', etc. Gallego (2022, p. 299) refers to these groups as the consist type (i) and the insist type (ii).The verbs discussed in this paper behave like the insist type, as also noted by Casanova (2020), i.e., they admit the four context considered.Needless to say, there are differences with respect to membership in each group, according to language variation.12 Although we do not discuss explicitly the SO∼AO with adjectives, the analysis we propose applies also in this domain.We go back to this in the following subsection.

13
For expository reasons we are simplifying the structure for nosotros.In the case of Spanish nosotros 'we', a complex form is recognized when comparing it with other romance languages: nos-otros vs. Italian noi, French nous; Portuguese nós.Nos clearly lexicalizes the information related to 1PL, as can be observed in the correspondent clitic nos 'us', while otros 'others' presents all the characteristics of an adjetive-like element merged in Spec nP.From that position, the n[uϕ] of it copies deictic gender and number information from the main pronominal structure.It means that a refined (33) structure should have the projections for otros 'others' and nos would be the lexical item for the main structure.

14
We propose that there would be three types of ns: (1) with unvalued ϕ-features, which would be adjective-like ns; (2) n without features in a structure that presents only deictic information; and (3) n with arbitrary features that, in the case of Spanish, are generally arbitrary gender features in non-animate nouns.Basically, these would be the three ways in which noun-like elements would be individualized in terms of Borer (2005).See Estomba (2017) for more detailed discussion on the relation between gender and other projections in the DP structure.We leave aside adjectives for now.

16
In fact, the presence of SOs with adjectives seems to be the most restrictive option.Speakers who use the SO with verbs productively do not necessarily employ it with adjectives, but the reverse is true.This could be due to the fact that the adjective itself has a n with unvalued features, so that the defectivity is not as clearly defined as in adverbs that has a defective n or in verbs in which we do not even find a n.

Figure 1 .
Figure 1.Grammar Structure in DM (PF for Phonological Form, LF for Logical Form).

Figure 1 .
Figure 1.Grammar Structure in DM (PF for Phonological Form, LF for Logical Form).
proposal.However, I partially take the Nanosyntax view on features as projections and, consequently, in the system developed here Hum, Part, and Auth project in the structure.The projection PartP can be positively [+] or negatively [−] marked, while AuthP can be absent in the structure.HumP can merge with both Auth and Part 10 .The feature [+PL] on # can be the result of two PartP with opposite features or of AuthP and PartP in #'s c-command domain.Given the above, the basic structures for pronouns are schematized below.are materialized by different Vocabulary Items.A Vocabulary Item (VI from now on) is understood as a relation between syntactic-semantic features/parts of structure and phonological exponents.When Vocabulary Insertion takes place, it is said that a specific part of the structure has been lexicalized.The scheme in (16) represents a VI in terms of the Distributed Morphology framework.(16) /phonological exponent/↔ [syn-sem features] Languages 2023, 8, x FOR PEER REVIEW 8 of 20 (13) 1PL inclusive (14) 1PL exclusive (15) 2PL exclusive These abstract structures are materialized by different Vocabulary Items.A Vocabulary Item (VI from now on) is understood as a relation between syntactic-semantic features/parts of structure and phonological exponents.When Vocabulary Insertion takes place, it is said that a specific part of the structure has been lexicalized.The scheme in (16) represents a VI in terms of the Distributed Morphology framework.(16) /phonological exponent/↔ [syn-sem features] Languages 2023, 8, x FOR PEER REVIEW 8 of 20 (13) 1PL inclusive (14) 1PL exclusive (15) 2PL exclusive These abstract structures are materialized by different Vocabulary Items.A Vocabulary Item (VI from now on) is understood as a relation between syntactic-semantic features/parts of structure and phonological exponents.When Vocabulary Insertion takes place, it is said that a specific part of the structure has been lexicalized.The scheme in (16) represents a VI in terms of the Distributed Morphology framework.
These abstract structures are materialized by different Vocabulary Items.A Vocabulary Item (VI from now on) is understood as a relation between syntactic-semantic features/parts of structure and phonological exponents.When Vocabulary Insertion takes place, it is said that a specific part of the structure has been lexicalized.The scheme in (16) represents a VI in terms of the Distributed Morphology framework.(16)/phonological exponent/↔ [syn-sem features] Languages 2023, 8, x FOR PEER REVIEW 8 of 20 (13) 1PL inclusive (14) 1PL exclusive (15) 2PL exclusive These abstract structures are materialized by different Vocabulary Items.A Vocabulary Item (VI from now on) is understood as a relation between syntactic-semantic features/parts of structure and phonological exponents.When Vocabulary Insertion takes place, it is said that a specific part of the structure has been lexicalized.The scheme in (16) represents a VI in terms of the Distributed Morphology framework.(16) /phonological exponent/↔ [syn-sem features] Languages 2023, 8, x FOR PEER REVIEW 8 of 20 (13) 1PL inclusive (14) 1PL exclusive (15) 2PL exclusive These abstract structures are materialized by different Vocabulary Items.A Vocabulary Item (VI from now on) is understood as a relation between syntactic-semantic features/parts of structure and phonological exponents.When Vocabulary Insertion takes place, it is said that a specific part of the structure has been lexicalized.The scheme in (16) represents a VI in terms of the Distributed Morphology framework.(16) /phonological exponent/↔ [syn-sem features] These abstract structures are materialized by different Vocabulary Items.A Vocabulary Item (VI from now on) is understood as a relation between syntactic-semantic features/parts of structure and phonological exponents.When Vocabulary Insertion takes place, it is said that a specific part of the structure has been lexicalized.The scheme in (16) represents a VI in terms of the Distributed Morphology framework.(16)/phonological exponent/↔ [syn-sem features] (17) Indo-European adjectives [Adapted from Mitrović and Panagiotidis 2020, p. 11] [Adapted from Mitrović and Panagiotidis 2020, p. 11] This bicategorial property explains that Indo-European adjectives are modified by adverbs-like verbs-and present nominal features like gender and number by ϕ-agreement (concord).
) proposes argument configurations for verbs, which are the result of combining relational projections (VoiceP, PathP, and PlaceP) with non-relational elements (DPs and Roots).Place establishes a predicative relation, Path encodes a transition, and Voice introduces the Originator of the event.Roots and DPs are interpreted according to the position they occupy in the structure.For instance, a DP in the scope of VoiceP can only be interpreted as an Agent or a Cause.The DPs in the domain of lower heads are interpreted as Themes or Locations depending on the relation with the head they merge with.The configuration in (18) represents the structure proposed byAcedo-Matellán (2016, p. 35)  for transitive events of change of state/location, with v as a verbalizer according toMarantz's (1997) Categorization assumption.(18) a. b. guments arise from their locus in the syntactic structure.Acedo-Matellán (2016) proposes argument configurations for verbs, which are the result of combining relational projections (VoiceP, PathP, and PlaceP) with non-relational elements (DPs and Roots).Place establishes a predicative relation, Path encodes a transition, and Voice introduces the Originator of the event.Roots and DPs are interpreted according to the position they occupy in the structure.For instance, a DP in the scope of VoiceP can only be interpreted as an Agent or a Cause.The DPs in the domain of lower heads are interpreted as Themes or Locations depending on the relation with the head they merge with.The configuration in (18) represents the structure proposed by Acedo-Matellán (2016, p. 35) for transitive events of change of state/location, with v as a verbalizer according to Marantz's (1997) Categorization assumption.(18) a. b.Some cartographic studies, such as Fábregas (2007a), Pantcheva (2011), or Gibert Sotelo (2017), propose more complex structures for locative interpretations based on crosslinguistic variation.Pantcheva, for instance, proposes that notions like Source as well as guments arise from their locus in the syntactic structure.Acedo-Matellán (2016) proposes argument configurations for verbs, which are the result of combining relational projections (VoiceP, PathP, and PlaceP) with non-relational elements (DPs and Roots).Place establishes a predicative relation, Path encodes a transition, and Voice introduces the Originator of the event.Roots and DPs are interpreted according to the position they occupy in the structure.For instance, a DP in the scope of VoiceP can only be interpreted as an Agent or a Cause.The DPs in the domain of lower heads are interpreted as Themes or Locations depending on the relation with the head they merge with.The configuration in (18) represents the structure proposed byAcedo-Matellán (2016, p. 35)  for transitive events of change of state/location, with v as a verbalizer according toMarantz's (1997) Categorization assumption.(18)a.b.Some cartographic studies, such asFábregas (2007a),Pantcheva (2011), or Gibert Sotelo (2017), propose more complex structures for locative interpretations based on crosslinguistic variation.Pantcheva, for instance, proposes that notions like Source as well as Some cartographic studies, such asFábregas (2007a),Pantcheva (2011), or Gibert Sotelo (2017), propose more complex structures for locative interpretations based on crosslinguistic variation.Pantcheva, for instance, proposes that notions like Source as well as Goal project in the structure hierarchically.In (19), the item to would lexicalize both Place and Goal, while in (20) from lexicalizes Place, Goal, and Source.(19) To the house (20) From the house Languages 2023, 8, x FOR PEER REVIEW 10 of 20

(
35) a. Una amiga tuya (ok for all Spanish varieties) A friend of yours b. *Cerca tuyo (* for some Spanish varieties) Near of you c. *Me acordé tuyo (* for most of Spanish varieties) I remembered you Languages 2023, 8, x FOR PEER REVIEW 14 of 20

(
35) a. Una amiga tuya (ok for all Spanish varieties) A friend of yours b. *Cerca tuyo (* for some Spanish varieties) Near of you c. *Me acordé tuyo (* for most of Spanish varieties) I remembered you 6 4. (i) Las paredes son alt-a-s.The.F.PL wall.F.PL are high-F-PL.'The walls are high'.(ii) Las aves vuelan alt-o.The.F.PL bird.F.PL fly high-DEF.'Birds fly high'.8 11 15

Table 1 .
Synthetic and analytic forms.
de + nonagreeing syntactic objects (examples in b) are in complementary distribution.

Table 5 .
Valuation for n [uϕ]in different domains and variation.
; Marttinen Larsson and Bouzouita 2022, for further discussion).The three examples are translated as 'near you'.