Dative experiencers and (null) subjects in Peninsular Spanish infinitives

Dative experiencers have been argued to have certain subject properties in finite domains. The question thus arises whether they can ever be controlled like structural (null) subjects. In the literature, it has been argued that one feature distinguishing dative experiencers from (nominative) subjects in Spanish is that the former cannot be controlled, differently from true quirky subjects. By examining corpus data, I argue that Spanish dative experiencers, even though they cannot be obligatorily controlled in complement infinitives, can appear in adjunct infinitives in non-obligatory control contexts. One property that is crucial for sanctioning this option is the possibility of licensing full DP subjects in nonfinite domains. If the subject position is occupied by a non-controlled nominative DP, dative experiencers are bound by logophoric or topic coordinates in C. The data give further support to an Agree -based theory of control, according to which the referential relation between the subject of infinitives and its controller is mediated by functional heads of the extended verbal projection.


Introduction
Dative experiencers have been observed to share some (but not all) properties of subjects in Spanish (cf. Masullo 1993, Fernández Soriano 1999, Gutiérrez-Bravo 2006, Cuervo 2010, Fábregas et al. 2017. For example, SVO is the unmarked word order with transitive verbs in Spanish, being compatible with an all-focus interpretation. With psych verbs, however, it is the dative experiencer, and not the nominative subject, that is in preverbal position in all-focus sentences (Fernández Soriano 1999, Fábregas et al. 2017): (1) A: Qué pasa? 'What's up?' B1: (Que) a Pedro le molesta el humo.
that to Pedro him bothers the smoke 'Peter is bothered by the smoke.' B2: # ( Que) le molesta el humo a Pedro.
to Lucy her.DAT like Ronny before know to Otto 'Lucy liked Ronny before getting to know Otto.' (Campos 1999(Campos : 1560 (3) a. Al príncipei le gustó Soraya para sí mismoi. to.the prince CL liked Soraya for himself 'The Prince liked Soraya (to keep her) for himself.' b. * Marta le habló al psiquiatra i de sí mismo i . Marta CL spoke to.the psychiatrist about himself (Gutiérrez-Bravo 2006: 2, citing González 1988 (4) Le ocurrió un accidente borracha. her.DAT happened an accident drunk 'An accident happened to her while being drunk.' (Fernández Soriano 1999: 124) Some authors have taken the similarity of dative experiencers and subjects as evidence for the assumption that Spanish dative experiencers can occupy the subject position Spec,IPa position with mixed A-and A'-properties (cf. Masullo 1993).
However, other authors have stressed the differences between dative experiencers and nominative subjects as well as quirky subjects of the Icelandic type (see e.g., Gutiérrez-Bravo 2006, Fábregas et al. 2017. One argument against a quirky subject analysis comes from nonfinite control structures. Thus, Masullo (1993) observes that Spanish dative experiencers, differently from Icelandic quirky subjects (see e.g., Zaenen et al. 1985), cannot be controlled, i.e., they cannot be PRO: (5) * Yo espero PRO agradarme el concierto. I hope.1SG like.INF.ME.DAT the concert 'I hope to like the concert.' (Masullo 1993: 309) Similar arguments have been discussed by Gutiérrez-Bravo (2006) and Fábregas et al. (2017). However, it is an open question whether the ungrammaticality of structures like (5) derives from the impossibility to control dative experiencers per se or from the impossibility of a nominative DP subject inside the control infinitive, given that the DP el concierto would need nominative Case. Thus, even though nominative DP subjects are possible in the personal infinitive in Spanish, they are ruled out in complement control infinitives like (5) (see Piera 1987, Hernanz 1999. In this paper, I have a look at dative experiencers in nonfinite control contexts and I argue that the generalization that dative experiencers cannot appear in control contexts needs to be refined. While they cannot be controlled in complement infinitives like (5), dative experiencers show patterns of highly preferred co-reference with a matrix antecedent in some Spanish adjunct infinitives, as in the following sentence: (6) [...] dijo Amenábar, un " agnóstico", quieni no hizo esta película said Amenabar an agnostic who not made this movie [por interesarlei en particular el tema de la eutanasia]. for interest.INF.him.DAT particularly the topic of the euthanasia '[…] Amenabar said, an 'agnostic' whoi didn't make this movie because hei was particularly interested in euthanasia.' (CORPES XXI,written,Spain) 1 In fact, the dative experiencer of interesar 'interest' has quite similar 'control' properties as the null subject of the corresponding reflexive verb interesarse (por) 'to be interested in': (7) [...] quieni no hizo esta película [por i interesarse por el tema who not made this movie for interest.INF.REFL for the topic de la eutanasia]. of the euthanasia Note that the adjunct infinitive in (6), apart from allowing a dative experiencer that is apparently controlled by a matrix antecedent, also sanctions the nominative DP subject el tema de la eutanasia 'the topic of euthanasia'.
It has been suggested that complement control structures like (5) exhibit Obligatory Control (OC), while adjunct infinitives allow both, Obligatory and Non-Obligatory Control (NOC; in the sense of Williams 1980Williams , 1992. In OC structures, the 1 All emphasis markings, indexes, translations and glosses that appear in corpus examples from CORPES XXI in this paper have been added by myself. reference of the null subject of the infinitive must obligatorily be recovered from a ccommanding and local (i.e., matrix) antecedent. In NOC structures, on the contrary, the null subject does not need an antecedent (i.e., it can receive an arbitrary interpretation) or it can be identified in a non-local relation (i.e., long distance). In fact, it has been argued that the null subject in NOC structures is either logophorically controlled (see Williams 1992, Landau 2013 or it is a pronominal pro (cf. Rigau 1995, Torrego 1998 for Catalan and Spanish: see also Hornstein 1999).
In this paper, I argue that the possibility of NOC dative experiencers depends on the licensing of overt, nominative DP subjects and pro in Spanish adjunct infinitives (Hernanz 1982, Rigau 1995, Mensching 2000, Herbeck 2015. The data provide additional evidence for the assumption that the null subject in Spanish adjunct control infinitives is not homogeneously PRO, but it can be similar to pro in certain contexts. However, given the lack of pronominal agreement, this null subject needs to be directly bound by logophoric or topic coordinates in the left periphery. This yields a strong tendency of (null) subjects as well as dative experiencers to be (locally) co-referent with a matrix antecedent. The analysis will be supported by a detail analysis of data from the sub-corpus 'Spain' of CORPES XXI (RAE).
This paper is structured as follows: first, I discuss the theoretic background with respect to obligatory control, non-obligatory control, the possibility of overt DP subjects, and the role of dative experiencers in Spanish infinitives. Thereafter, I outline the study of data from the corpus CORPES XXI (RAE), containing sentences with an adjunct infinitive and an apparently controlled dative experiencer. In section 4, I provide a syntactic analysis of dative experiencers and (null) subjects in Spanish infinitives. In section 5, some cross-linguistic remarks will be made. Section 6 is dedicated to the conclusion.

Background: PRO, pro and overt DP subjects in Spanish infinitives
The difference between OC and NOC is well established ever since Williams (1980). As has been pointed out, the null subject of OC infinitives must have a c-commanding and local antecedent in the next higher matrix clause (see also Hornstein 1999, Landau 2000. NOC infinitives, on the other hand, can have null subjects that receive an arbitrary interpretation or that are controlled by a discourse antecedent (see Landau 2013Landau , 2015 Landau 2000Landau , 2013 for discussion and further references). In this paper, I rely on (written and oral) corpus data so that I can only make observations with respect to the diagnostics of locality and c-command.
In Spanish, the difference between OC and NOC is complemented by a further option: adjunct infinitives allow the subject position to be occupied by a nominative subject DP or pronoun (see Hernanz 1982, Rigau 1995, Torrego 1998, Mensching 2000, Pöll 2007, Herbeck 2015: 3 (10) De abrir Julia la puerta tendremos que marcharnos. of open.INF Julia the door will.have.to.1PL go.away.REFL1PL 'If Julia opens the door, we will have to go away.' (Piera 1987: 164) It has been argued that adjunct infinitives in the languages Catalan and Spanish have 'abstract AGR' on T given that, apart from allowing nominative DP subjects, they allow non-controlled null subjects in some contexts, differently from subject infinitives, which are more restrictive (see Rigau 1995for Catalan, Hernanz 1999 for Spanish): (11) * Fue penoso desmayarte en aquel lugar. was embarrassing faint.INF.REFL.2SG in that place (Hernanz 1999(Hernanz : 2267 (12) Al desmayarte, empezaron a chillar. in.the faint.INF.REFL.2SG began.3PL to shout 'When you fainted, they began to shout.' (Rigau 1995: 286) As the contrast between (11) and (12) shows, subject infinitives do not easily allow free reference of their null subject, while adjunct infinitives are more permissive.
Furthermore, adjunct infinitives allow null expletives in Spanish (see Torrego 1998, Herbeck 2021 In the literature, these data have been taken as evidence that adjunct infinitives in Spanish can license pro (Rigau 1995, Torrego 1998, Herbeck 2021. In OC contexts, on the other hand, overt NP subjects (definite and indefinite ones) as well as null subjects with free reference are out in Spanish, even though emphatic pronouns and quantifiers are possible if they are co-referent with a ccommanding matrix antecedent (cf. Piera 1987, Mensching 2000, Herbeck 2015: 4 3 The possibility of full DP subjects is more restricted in subject infinitives than in adjunct infinitives (Piera 1987, Hernanz 1999. The present paper focuses on the latter structure. 4 There are different theoretical proposals with respect to the analysis of structures like (14): Piera (1987) argues in favor of an approach in which the emphatic pronoun is an adjoined element and doubled by PRO in subject position: (14) Julia i prometió [hacer (ella misma i/*x / *Juan) la cena]. Julia promised make.INF she self Juan the dinner 'Julia promised to prepare the dinner herself / *John.' Thus, in OC contexts, phonetically realized subjects are only possible if they are [-R] in the sense of Reinhart & Reuland (1993) and Landau (2000Landau ( , 2004, i.e., if they are phi-deficient and referentially dependent elements, but not if they are [+R], i.e., if they carry a full set of phi-features and Case, such as (free) pronouns and lexical NPs. 5 Even within the group of adjunct infinitives, the productivity of [+R] subject licensing is not homogeneous, but it depends on various factors, such as the type of introducing preposition, the position of the infinitive (extraposed vs. intraposed), and the degree of (non-)integration (in the sense of Haegeman 2012) of the adjunct (see Vanderschueren 2013, Herbeck 2021 for corpus analyses; see also Fernández Lagunilla & Anula Rebollo 1994, Hernanz 1999, Mensching 2000, Pérez Vázquez 2007 for discussion). For example, Galán Rodríguez (1999: 3621) notes that the null subject of a para-infinitive with a prospective meaning, implying an intentional agent, must be co-referent with a matrix antecedent: 6 (15) Robó una pistola para atracar un banco. (Hernanz 1999: 2313) '(S)he stole a gun in order to rob a bank.' However, as noted above with respect to (12), several adjunct infinitives allow non-control of their null subject. Even with the same preposition para 'for', 7 noncontrolled null subjects can be found. This is the case if the para-infinitive has a concessive interpretation (Sánchez López 1995) and it is in a peripheral position (see Herbeck 2021): (16) […] en Madridi la policía yo creo que sí que trabaja bien // in Madrid i the police I think that yes that work.3SG well Other authors suggest that 'overt PRO' exists in Spanish (see Mensching 2000, Alonso-Ovalle & D'Introno 2001, Herbeck 2015 and several other languages (see Szabolcsi 2009, Landau 2013, and references therein, among others). 5 In Reinhart & Reuland (1993), phi-feature deficiency is translated into referential dependence, i.e., [-R]: SE and self-anaphors are referentially dependent [-R] elements (lacking a full set of phi-features) and pronouns are referentially independent [+R] elements (cf. ibid. 659), having a full set of phi-features (and Case). Landau (2000Landau ( , 2004 extends this system to PRO, arguing that it is a phi-deficient [-R] element. In Herbeck (2015), it is argued that emphatic pronouns as in (14) in Spanish can thus be analyzed as morpho -syntactically phideficient D[ɸ:_] elements, which are phonetically realized post-syntactically as a pronoun after phi-feature valuation via Agree. 6 A reviewer notes that this example would also allow Partial Control, i.e., the null subject of atracar 'to rob' could be the matrix antecedent plus another referent. However, as we will see below (examples (17) and (30)), this type of para-infinitive does not allow apparently controlled dative experiencers, similarly to other OC infinitives. This is expected if Partial Control is a subtype of OC and not NOC (see Landau 2000Landau , 2013. 7 For further discussion of the referential and semantic properties of different types of infinitives and subjunctives with para in Spanish, see Morales (1989 In (16), the null subject of the peripheral concessive para-infinitive does not have a c-commanding antecedent.
The data with Spanish infinitives thus show that the referential dependency of a null subject inside an infinitive on an antecedent is a matter of degree, ranging from full referential dependence (complement control) to referential independence (nonintegrated adjunct infinitives), with integrated adjunct infinitives being in between.
In the next section, we will see that these properties of Spanish adjunct infinitives are crucial to understand the behavior of dative experiencers in infinitives.

Dative experiencers in Spanish adjunct infinitives
As has been mentioned in section 1, dative experiencers cannot be controlled in complement infinitives (see (5)). In Gutiérrez-Bravo (2006), it is further argued that the dative experiencer cannot be controlled in the following adjunct infinitive: As has been noted in (6), however, dative experiencers can be co-referent with a matrix antecedent in some Spanish adjunct infinitives. The following example demonstrates the same phenomenon in an infinitive introduced by sin 'without': (18) cada unoi mira por lo suyo, [ sin importarlei los demás] each one looks for the own without matter.INF.CL.DAT the others 'Everybody looks after his own [people], without caring for the others.' (CORPES XXI, written, Spain) Note that the para-infinitive in (17) implies a purpose and there is a relation of intentionality between the matrix agent and the embedded event. These contexts require control of the null subject according to Galan Rodríguez (1999) and Hernanz (1999) (see (15); cf. also Gómez et al. 2022 for an OC analysis of some parainfinitives). Thus, it is OC, but not NOC, that blocks local co-reference of dative experiencers with a matrix antecedent in complement as well as adjunct infinitives.
In the next subsections, we will have a look at (i) further evidence which reinforces the parallelism between dative experiencers and null subjects in Spanish NOC adjuncts and (ii) the contexts in which apparent NOC dative experiencers are licensed.

The data
All data on which the discussion of dative experiencers in infinitives is based stem from the corpus CORPES XXI (RAE), which contains texts from various sources of different varieties of Spanish (Spain and Latin America) and from different registerswritten as well as oral. In this paper, I restrict the data to the sub-corpus of Spain because it is well-known that Latin American varieties show some differences from Peninsular Spanish with respect to the licensing of overt subjects in (adjunct) infinitives (see Suñer 1986), which might affect the phenomenon investigated here.
To extract the data, I looked for nonfinite psych verbs with a 3 rd person dative cliticin particular, gustar 'like', encantar 'to love', importar 'matter', interesar 'interest', molestar 'bother', preocupar 'worry'in the context of an introducing preposition. The sentences have been analyzed in detail with respect to (i) the type of the introducing preposition, (ii) (co-)reference properties of the dative experiencer of the infinitive (locally or non-locally identified), and (iii) the type of the nominative subject (null or overt and, if null, whether locally or non-locally identified). With respect to (ii) and (iii), I consider a dative experiencer or a (null) subject inside an adjunct infinitive to be locally identified (see (19)) if it has an antecedent in the matrix clause which is not embedded inside another constituent (i.e., if it potentially ccommands an adjunct clause) and to be non-locally identified otherwise (i.e., if the matrix antecedent is embedded inside another constituent or if it is not mentioned in the main clause; see (20)): In total, 268 adjunct infinitives including a psych verb have received a detailed analysis. I previously excluded all infinitives introduced by a preposition which were verbal periphrasis, such as raising, aspectual and modal constructions (empezar a, comenzar a 'start', deber de 'ought to', ir a 'going to', etc.). As becomes clear from Tables 1 and 2, there is a high frequency of the preposition sin 'without' and the psychverb importar 'care/matter' in the sample examined here:  In fact, 229 out of the 268 analyzed adjunct infinitives (= 85%) were the combination sin + importar 'without caring; regardless of', which might point to a certain degree of fixation of this sequence: Even though rarer in the corpus data, examples with other prepositions and verbs exist, as for e.g., por 'for' + interesar 'interest' in (6), aparte de 'beside' + encantar 'excite' in (24) or para 'for' + gustar 'like/please' in (31) below, among others.
In the examined configuration, local identification of the embedded dative experiencer with a matrix antecedent typically arises and the structural subject of the infinitive is often realized as a [+R] DP or as a null expletive associated with a postverbal que-clause. This issue will be further discussed in the next section.

Dative experiencers and null subjects in Spanish adjunct infinitives: How similar are they?
Examples like (6), (18) and (21) show that dative experiencers can be co-referent with a matrix antecedent in several NOC infinitives, differently from OC ones. However, this still does not demonstrate that they behave similarly to (null) subjects in these configurations. This section lays out some evidence that dative experiencers and null subjects behave similarly in Spanish adjunct infinitives with respect to their coreference properties. However, it will also be shown that neither of the two elements should necessarily be defined as obligatorily controlled. I conclude that dative experiencers have the possibility to appear in NOC contexts, which is made possible by the Spanish grammar by the availability of (null and overt) non-controlled subjects in infinitives.
Some evidence in favor of the assumption that dative experiencers behave similarly to null subjects with respect to their identification comes from coordination: In (22), the adjunct infinitive introduced by sin 'without' contains a psych verb with the dative experiencer le 'to him', which is locally co-referent with the ccommanding matrix null subject of había dicho '(s/he) had said'. This adjunct infinitive is coordinated with an infinitive whose null subject locally co-refers with the same matrix antecedent, indicating that the referential properties of the dative and (null) subject are similar in this configuration.
A similar situation obtains in (23), where two adjunct infinitives introduced by sin 'without' are coordinatedone with a dative experiencer and the other one with a null subject. Again, both elements are linked to the same matrix controller, indicating that co-reference properties are similar for the two elements: 8 In several of the previous sentences, the dative experiencer is the only element of the non-finite psych verb that could be 'controlled' by a matrix antecedent because the structural (nominative) subject is either overtly realized as a full nominal element as in (6), (18) or (23), or as a CP-clause as in (21), arguably associated with a null expletive in Spec,TP.
If the subject is not an overt NP or a clause, the null subject and the 3 rd person dative experiencer of the psych verb compete for 'control' by a matrix antecedent.
(24) A mi chicoi, [ In (24), both elementsthe null subject and the dative experiencer clitic of the nonfinite psych verbare locally identified by a matrix antecedent. What is interesting is that, in other cases of competition, the structural null subject is not locally controlled, but the dative experiencer is: 8 A reviewer asks what it is that prevents a structure of sentences like (23) with PRO, i.e., "sin PROi importarlei". Such an analysis would be problematic because the subject position can be occupied by a nominative NP that is not coindexed with the dative experiencer (see e.g., example (18)). Below it is argued that Spec,TP is either occupied by expletive pro or it is fully absorbed in the relevant configurations. Here, the dative experiencer co-refers with the subject of the next higher matrix clause, while the null subject of the infinitive is non-locally identified by the first person singular null subject of comprendí '(I) understood' or the non-c-commanding possessive pronoun mi 'my'.
In the next section, I descriptively present the frequency of dative experiencers in Spanish adjunct infinitives entering local or non-local identification in the examined sample.

Preferred but not obligatory local identification
If we look at the frequency of local identification (i.e., co-reference with a matrix antecedent that is not embedded inside another constituent) and non-local identification of dative experiencers inside adjunct infinitives, it becomes clear that the former is strongly preferred in the examined sample. However, non-local identification is possible so that OC is not an option in these cases: In Table 3, out of 264 weak (i.e., non-doubled) dative experiencers in nonfinite adjunct infinitives (4 of the 268 infinitives contained a strong pronoun or DP experiencer), 95% are identified by a matrix antecedent. This tendency mirrors the situation of null subjects which, according to data in Herbeck (2021), are preferably subject to local matrix control (81%), but non-local identification is possible. The strongest tendency of local identification can be observed with the psych verb importar 'matter' (98%), while other psych verbs, which are also far less frequent in the sample, show more variation (79% local).
Note that, similarly to the subject position of adjunct infinitives, insertion of a strong pronoun or DP experiencer is structurally possible, even though very rare (4/268 = 1,5%): (26) estás lo suficientemente bueno como para gustarle a Mónica is.2SG the sufficiently good as for like.INF.her.DAT to Mónica 9 I use the term "weak" dative experiencer for a configuration that only contains a dative clitic, but not a double in the form of a strong pronoun or DP (i.e., sin importarle vs. sin importarlei {a éli / a Juani}), in parallel to configurations with a null subject vs. overt pronoun or subject DP (i.e., sin Ø saberlo vs. sin saberlo {él / Juan}).
'you are good looking enough so as to be liked by Monica' (CORPES XXI, written, Spain) Furthermore, I have only found such elements in postverbal, but not in preverbal, position in the data. This mirrors the behavior of the subject position in adjunct infinitives which is empty in the unmarked case but can be occupied by a postverbal strong pronoun or DP (see e.g., (10)). 10 Let us have a look at the referential properties and the (overt or null) realization of nominative subjects with non-finite psych verbs: Similarly to dative experiencer clitics, null subjects of psych verbs have a strong tendency to be locally identified (90%), i.e., to be co-referent with a c-commanding matrix antecedent. The frequency of overt DPs or clausal subjects is very high with the verb importar 'matter' (99%), which again indicates a special behavior of this verb in nonfinite domains. Other psych verbs also have a high frequency of 29% of overt subjects, considering that nonfinite domains usually have a lower rate of expressed subjects (3.5% and 5% in two corpora of the study in Herbeck 2021). I take these descriptive patterns to indicate that local identification of dative experiencers in nonfinite domains is above all favored in contexts in which the structural subject is overtly realized either as a [+R] DP or as a clause, as is the case with importar 'matter'. Other psych verbs with less frequent subject expression also have more variable patterns with respect to the identification and co-reference patterns of dative experiencers.
For all psych verbs, it can be concluded that dative experiencers do not show patterns of OC, but of co-reference (in the sense of Reinhart 1983) and there is a strong preference for it to hold with a matrix antecedent, which is similar to null subjects in other Spanish adjunct infinitives. The next section provides some further evidence for this reasoning from phi-agreement patterns and (lack of) c-command.

NOC of dative experiencers and null subjects in Spanish adjuncts
Even though 3 rd person dative experiencer clitics have a strong tendency of being locally identified, above all with the verb importar 'matter' (98%), it has been argued that we are not dealing with obligatory control. This is reconfirmed by a detail analysis of several sentences. First, c-command is not a categorical requirement on the referential dependency between the dative experiencer and the antecedent: 10 It has been observed in the literature on Peninsular Spanish control infinitives that overt subjects are preferably in postverbal position, the preverbal position being available in adjunct infinitives for some speakers, but only very restrictedly (see Hernanz 1999, Mensching 2000, Pöll 2007).
(27) El director i es Dios y, por ello, sui trabajo decide la vida de sus the director is God and therefore his work decides the life of his súbditos (personajes) sin importarlei la opinión del público. subjects characters without matter.INF.to.him the opinion of.the audience 'The director is God and, therefore, his work decides the life of his subjects (characters) regardless of the opinion of the audience.' The following example demonstrates that the null subject of a psych-verb, similarly to dative experiencers, does not need a c-commanding antecedent: 11 (28)  What is interesting in this example is that, apart from a lack of c-command, the antecedent of the structural null subject of gustar 'like' is [-human]. This way, (28) is neither a case of predicative nor of logophoric control, but of pragmatically determined co-reference, as is the case with pro (see Herbeck 2021 for further evidence). While the null subject is not locally identified by a c-commanding matrix antecedent, the dative experiencer le 'to him' in the same sentence is. This indicates again that (null) subjects and dative experiencers compete for local matrix identification in adjunct infinitives containing a psych verb. As I argue below, the dative experiencer has more proto-typical properties for topic as well as logophoric identification than the null subject and, given that both concepts are decisive for NOC (see Landau 2013, Herbeck 2021, dative experiencers are chosen over the null subject for local matrix identification in several examples. Note, furthermore, that an OC analysis of dative experiencers is ruled out also for phi-agreement purposes. In OC, a minimal pronoun (Kratzer 2009) acquires phifeatures from a c-commanding matrix antecedent either via predication or logophoric linking (Landau 2015). In adjunct infinitives, however, null subjects (see Herbeck 2021), as well as dative experiencers enter semantic (i.e., ad sensum) rather than morpho-syntactic agreement: (29) Y la gentei comprará las acciones para venderlas luego and the people.SG will.buy.3SG the shares for sell.INF.them later con beneficio alto, sin conocer siquiera los cuadros, with benefits high without know.INF not.even the pictures 11 Hornstein's (1999) OC analysis of adjunct control would predict a c-command requirement which, however, does not hold for Spanish.

sin
importarlesi la calidad artística […] without matter.INF.to.them the quality artistic 'And people(SG.) will buy the shares and sell them afterwards at a high profit, without even knowing the paintings, without caring about the artistic quality.' (CORPES XXI, written, Spain) In this example, the dative clitic could not be analyzed as a minimal (phi-less) pronoun which acquires phi-features through a syntactic Agree operation, given the number mismatch between singular la gente 'people' and the plural dative experiencer les. This way, the dative experiencer can be generated in syntax already specified for phi-features (just like pro, but not PRO) and enter co-reference patterns depending on various discourse parameters (in the sense of Reinhart 1983). 12

Not all adjunct infinitives allow NOC dative experiencers
In the preceding sections, I have argued that dative experiencers are similar to null subjects in adjunct infinitives in that both show patterns of preferred co-reference with a c-commanding matrix antecedent. Dative experiencers, however, lack the option of OC, as is demonstrated by various examples in the literature (Masullo 1993, Gutiérrez-Bravo 2006, Fábregas et al. 2017. Interestingly, also some adjunct infinitives block apparent non-obligatory 'control' of dative experiencers. In finality and purpose parainfinitives, locally co-referent dative experiencers are blocked, and control of the null subject arises: Note, however, that other types of para-infinitives do allow co-referent dative experiencers with a matrix antecedent but, in this case, there is no intentional relation between the matrix and the embedded events: (31) pues para no gustarlei la medicina como me dijo al principio well for not like.INF.to.him the medicine as me told at.the beginning / Øi sabe muchísimas cosas (she) knows many things 'Even though she doesn't like medicine, as she told me at the beginning, she knows a lot of things.' (CORPES XXI, oral, Spain) While in (30), the null subject of the infinitive is controlled by the matrix antecedent Julia and the dative experiencer has disjoint reference, it is the dative experiencer which is co-referent with a matrix antecedent in (31). As will be argued in section 4, 'local' matrix identification of the dative clitic in (31) correlates with a 12 Whether pro enters syntax specified for phi-features is a matter of debate (see Herbeck 2015 for an overview), given that it could be the pronominal agreement morpheme that is interpretable and assigns features to the null subject in finite clauses (see Rizzi 1982). However, for expository purposes, I adopt the notion of a phi-specified pro here. concessive interpretation of the infinitive and this type of infinitive is adjoined high in the structure, blocking OC and making available logophoric or topic control via left peripheral coordinates. At the same time, the nominative subject is overtly realized as a [+R] DP in the latter configuration.
In fact, in the majority of cases, if the dative experiencer is locally identified, the nominative subject is a non-controlled [+R] DP: out of the 252 locally identified dative experiencers in the sample, the subject was either an overt NP or a clause in 234 cases (93%). This indicates that matrix identification of dative experiencers in adjunct infinitives correlates to a high extend with the possibility of sanctioning non-controlled structural subjects in Spanish. This issue will receive a theoretic implementation in section 4.
3.6. Interim summary I have provided evidence that dative experiencers resemble null subjects in some Spanish adjunct infinitives. Evidence for a similarity between null subjects and dative experiencers in Spanish comes from high frequencies of local identification by a matrix antecedent and from coordination of infinitives containing a dative experiencer and a null subject with the same co-reference patterns. Furthermore, null subjects as well as dative experiencers can show semantic agreement (rather than morphosyntactic Agree) and they lack an obligatory c-command requirement for their antecedent. I have concluded that null subjects as well as dative experiencers are pronominal elements in Spanish adjunct infinitives, being identified through preferred co-reference with a matrix antecedent.

Analysis: Control via C in adjunct infinitives
The data regarding dative experiencers and null subjects with psych verbs are evidence for the following assumptions: Control into adjuncts is not always OC (pace Hornstein's 1999 analysis in terms of movement). In fact, infinitives introduced by sin 'without', which only allow subject control in English according to Hornstein (1999), have a strong tendency of sanctioning locally identified dative experiencers with psych verbs while, at the same time, the structural subject position is non-controlled (either a DP, a clause or pro). The observed patterns with psych verbs thus provide additional evidence for the assumption that pronominal null subjects are an option in Spanish adjunct infinitives (see Rigau 1995). The following subsections provide the technical implementation of (non-)control patterns in Spanish infinitives with psych-verbs.

OC, NOC and (null) subjects vs. dative experiencers
In Landau's (2000Landau's ( , 2013 Agree-based Theory of Control (ATC), complement OC is the result either of direct phi-feature transmission from a matrix antecedent to the embedded minimal pronoun (via syntactic Agree) or via indirect phi-transmission via anaphoric AGR on T or logophoric coordinates on C. Furthermore, according to Landau (2001), OC obtains as soon as the infinitive is inside the VP of the matrix predicate (either in the complement or specifier of V). This implies that also in some (but not all) subject infinitives, OC is triggered. 13 This way, OC obtains if the control infinitive is structurally integrated into the matrix VP domain.
Let us first have a look at how we can explain the impossibility of OC dative experiencers in Spanish in this theory. In Borer (1989), control is not triggered by inherent features of the null subject (i.e., PRO) but by properties of the verbal functional head it associates with. Thus, a null subject is controlled if it enters a relation with anaphoric (and not pronominal) agreement on INFL. In (32), I use the feature [iφ:self] (i.e., anaphoric, but interpretable phi-features) on T to depict Borer's (1989) anaphoric AGR. Through binding by [iφ:self] on T, the null subject Ø of the infinitive is linked to an antecedent in the matrix clause: Turning to psych verbs, the subject properties of dative experiencers with verbs of the gustar-type (the piacere-class in Belletti & Rizzi 1988) are explained by assuming that they are higher on a thematic hierarchy as the theme subject (see e.g., Gutiérrez-Bravo 2002, 2006 and, thus, they are first merged higher in the structure. In Cuervo (2010Cuervo ( , 2020 and Fábregas et al. (2017), the dative Experiencer is merged inside an Applicative Projection, which takes a stative vP as complement ((33) based on Cuervo 2010: 29): The theme DP in (33) receives nominative Case by T while the experiencer values its Case features internally to ApplP (see Cuervo 2010). 14 If infinitives containing a psych verb have the same internal structure below TP as finite clauses, it is expected that the dative clitic as the head of Appl can be doubled by an Experiencer pronoun or DP in the Spec of ApplP in postverbal position (see example (26)), after verb movement takes place: The fact that I haven't found any doubled experiencers in preverbal position in the data, similarly to restrictions on the position of overt nominative subjects in infinitives (see Rigau 1995, Hernanz 1999, could be explained by the defective left periphery of Spanish infinitives (e.g., Gallego 2010), which blocks left peripheral fronting operations.
However, if the OC analysis in (32) is applied to a structure containing a nonfinite psych verb, a non-trivial problem arises: if subject PRO is just a minimal pronoun which receives phi via Agree, the question arises why a non-doubled dative experiencer in Spec,Appl cannot enter syntax as a minimal pronoun, be bound by 13 However, NOC obtains with certain matrix predicates if the infinitive is removed from the matrix VP domain (see Landau 2001 for details). 14 In Fábregas et al. (2017), the experiencer is a defective intervener and it moves to Spec,FP so that it doesn't interfere in the relation between T and the subject. However, this reasoning is problematic for infinitives with psych verbs and nominative subjects given that experiencers do not move to preverbal position in these structures. anaphoric AGR and be obligatorily controlled. In fact, it is well known from languages like Icelandic that Case-marked PRO exists (see e.g., Sigurðsson 2008 and references therein) and, in Spanish, it has been argued that the subject position of OC infinitives is not obligatorily empty, at least if it is focused (Alonso-Ovalle & D'Introno 2001, Barbosa 2009, Herbeck 2015. 15 Furthermore, the dative experiencer in a structure like (33) would be a closer Goal for binding by anaphoric AGR than the null subject.
I would like to argue that the reason lies in the lack of nominative [+R] subjects in Spanish OC infinitives rather than in an impossibility of dative experiencers to be controlled per se: The dative experiencer could potentially be bound by anaphoric AGR, but this would leave the structural subject either without phi or without Case so that the derivation crashes. Consider first the scenario in which boththe theme subject and the dative experiencerenter syntax as a minimal, phi-less pronoun (in the sense of Landau 2015) and the experiencer is obligatorily controlled: 'Juan i promises that he k /she k will like him i .' This configuration converges because all elements can value their features: the experiencer clitic receives dative case inherently to the ApplP (see Cuervo 2010Cuervo , 2020 and enters syntax specified for phi. The minimal subject pronoun (D [_] ) is a possible Goal of Agree with anaphoric phi on T because stative vP is unaccusative and not a strong phase (see Cuervo 2010; see Chomsky 2001 for weak phases and long distance Agree). This way, the dative experiencer as well as the minimal D subject are potential targets, but only under Agree with subject D [_] does the derivation converge, given the necessity to value its phi-features.
Let us turn to a third option: the dative experiencer is a minimal pronoun, bound by anaphoric AGR and the theme subject D enters syntax fully specified for phi-features. This is the ungrammatical configuration discussed in most works, in which the subject is usually indicated as being overtly realized (see (5) and the following example from Fábregas et al. 2017): 15 According to the analysis of Spanish emphatic pronouns in Alonso-Ovalle & D'Introno (2001) and Herbeck (2015), PRO is just a minimal pronoun (in the sense of Kratzer 2009; see Landau 2015) and, thus, it can be overtly realized if associated with focus. Overt realization does not change the obligatory co-reference requirement with a matrix antecedent in OC contexts.
(37) * Espero ____ gustarme María. hope.1SG ____ like.me.DAT María 'Intended: 'I hope that I will like María.' (Fábregas et al. 2017: 41) In this configuration, the experiencer would be controlled via anaphoric AGR and the subject DP enters the derivation fully specified for phi-features. There are two possible ways to solve the ungrammaticality of (37): (i) given the presence of phifeatures, the subject would have to value its (nominative) Case features and nonfinite OC structures lack the possibility of nominative Case assignment (Chomsky 1981) or they assign null Case (Chomsky & Lasnik 1993). This reasoning could be motivated by the relation Bianchi (2003) draws between nominative Case assignment and the presence of an external logophoric center (which is responsible for deictic anchoring of the subject). Obligatory Control infinitives have a defective CP or a reduced FinP (Gallego 2010, Herbeck 2015 and, thus, lack an external logophoric center (Haegeman 2004) and nominative Case in Spanish so that a phi-specified subject cannot be licensed. Non-controlled dative experiencers are possible because they inherently value their case within the ApplP projection. (ii) nominative Case is available in Spanish infinitives, but a [+R] phi-specified subject is incompatible with [-R] anaphoric agreement (Landau 2004), i.e., T [φ:self] can bind a minimal, but not a full [+R] pronoun or DP. Both options successfully rule out OC of dative experiencers in Spanish, not because they cannot a priori be generated as minimal pronouns, but because Spanish OC infinitives cannot legitimate (nominative) [+R] subjects. This means that control of dative experiencers is potentially available, but OC is ruled out because of the requirements imposed on nonfinite structural subjects in Spanish.
With respect to adjunct control, on the other hand, two possibilities of control exist (Williams 1992, Landau 2013: OC (via predication) and NOC (via logophoric identification). The option between the two is governed by a variety of structural and pragmatic factors. Among the most important ones, NOC is favored if the infinitive is preposed (see Williams 1992, Fernández Lagunilla & Anula Rebollo 1994. Furthermore, it seems to be the case that the degree of (non-)integration (in the sense of Haegeman 2012 and related work) is decisive for control patterns in Spanish, depending both, on the type of preposition (Herbeck 2021) and the position of the infinitive.
The same reasoning that blocks co-referent dative experiencers in complement OC infinitives, blocks them in integrated OC adjunct infinitives. For example, finality or purpose para-infinitives block or strongly disfavor [+R] nominative subjects (Pérez Vázquez 2007) so that only a minimal D subject is sanctioned, which needs to value its phi-features via binding by the C-T spine. In fact, Gómez et al. (2022) argue that finality para-infinitives involve OC (and not NOC) with null and overt subjects: Laura has bought the book for she can.INF study.INF 'Laura has bought the book for herself to be able to study'. (Gómez et al. 2022: 3) We have seen in example (30) that 'controlled' dative experiencers are not possible in this type of para-infinitive but control of the null subject is enforced.
However, in the preposed para-infinitive in (31), which is separated from the main clause by a pause, matrix identification of the dative experiencer and, at the same time, overt realization of the subject as a nominative DP becomes possible. Note that the para-infinitive in (31) is not a purpose, goal or finality infinitive, but it adopts a concessive meaning (see Sánchez 1995, Hernanz 1999, Pérez Vázquez 2007. As has been discussed with respect to example (16) from spoken peninsular Spanish, null subjects with free reference are possible in this configuration.
In Herbeck (2021), it is argued that concessive para-infinitives like those in (31) imply the speaker's epistemic stance (cf. Galán Rodríguez 1999: 3620) and, therefore, the infinitive is directly adjoined high in a SA (Speech Act) phrase (in the sense of Cinque 1999, Speas & Tenny 2003 Other integrated infinitives (see e.g., (30)) are adjoined low within VP and, even though they can be fronted, they are integrated into the matrix VP domain at one point of the derivation: Thus, we crucially need to distinguish between VP-integrated infinitives (which can be fronted) and those which are directly adjoined high (see Haegeman's 2012 discussion of integrated and non-integrated adjuncts; see also Galán Rodríguez 1999 for Spanish) and, thus, they are not c-commanded by the matrix arguments at any point of the derivation.
From a peripheral position, OC via syntactic Agree is impossible and, thus, an alternative identification mechanism of the null subject is at stake. As has been mentioned, Williams (1992) and Landau (2013Landau ( , 2021 argue that apart from OC via predication, adjunct infinitives sanction logophoric control. Examples of non-ccommanding and [-human] referents in Spanish adjunct control of type (16) are further evidence that, apart from logophoricity, co-reference (in the sense of Reinhart 1983) and topic-linking are a possibility, i.e., differently from OC complement infinitives, adjunct infinitives have the additional option of topic anchoring in Spanish. 16 In this configuration, phi features on T, even though not pronominal, are not anaphoric either. In fact, Rigau (1995) and Torrego (1998) argue that Spanish adjunct infinitives contain 'abstract AGR', which I implement as interpretable [iφ:def] on T: Given the lack of pronominal [+R] agreement on T, 'minimal' null subjects cannot be assigned phi from T, but the subject must be directly C-linked. This yields a strong tendency of null subjects as well as dative experiencers to be 'locally' co- 16 Pérez Vázquez (2007) argues that adjunct infinitives in Spanish sanction an external logophoric center in the vein of Bianchi (2003), licensing overt subjects and nominative Case. referent with a matrix antecedent, 17 which is influenced by several scales at the syntaxpragmatics interface, such as thematic role and topicality hierarchies (see Givón 1983, Gutiérrez-Bravo 2002, 2006, Landau 2021, among many others). Topic-linking should be favored with agentive and animate pronominal (weak) elements, while postverbal [+R] subjects are favored in adjunct infinitives if they are non-agentive, [-human] and rhematic (cf. Vanderschueren 2013, Herbeck 2021 for discussion of corpus data).
I would like to argue that it is the same factors that govern apparent 'control' of a dative experiencer and a null subject in Spanish adjunct infinitives: with psych verbs, the structural subject has several non-prototypical subject (and topic) properties: it is non-agentivein fact, it is classified as an internal <theme> argument in several studies (cf. Belletti & Rizzi 1988, among many others). In the data examined here, it is frequently either [-animate] (see e.g., (29) and (31)) or a postverbal clause (see e.g., (21) and Table 4). This way, it is low on scales of topicality (cf. Givón 1983), making the dative experiencer the preferred option for establishing local co-reference. Furthermore, as has been mentioned, the dative experiencer is higher than the theme subject on thematic role hierarchies (cf. Gutiérrez-Bravo 2002, 2006. Lastly, logophoric linking is not only favored by [+human] referents but it requires it (cf. Landau 2013).
In the NOC configuration in (41), both, dative experiencers and (null) subjects are structurally candidates for C-linking so that a competition arises which is resolved by the before-mentioned factors and scales: In this configuration, both D in the specifier of ApplP or of vP are structurally available for binding by logophoric or topic coordinates in C: given that we are dealing with an unaccusative structure lacking an external argument, the vP is not a strong phase and the experiencer as well as the theme argument can be targeted. If the theme argument is realized as a [+R] DP or a clause, the dative experiencer is the only potential target of C-linking to a discourse antecedent. Furthermore, given that the experiencer is higher on scales of thematic roles and topicality than the theme subject, it is the preferred option for binding by C, explaining the high preference of local identification and the illusion of control.
To summarize, dative experiencers show apparent NOC patterns, similarly to null subjects in Spanish adjunct infinitives. This possibility exists because the dative experiencer in Spec,ApplP as well as the null subject can be directly anchored by topic and logophoric coordinates in C in Spanish: non-anaphoric, but defective AGR does not trigger obligatory control of the subject so that the latter can (but does not need to) be bound via C. Given that the experiencer is higher on scales of topicality and on thematic hierarchies than the theme subject, it is the preferred option for C-linking to a matrix antecedent in several configurations. 17 As has been shown in section 3, the tendency of 'local' co-reference with a matrix antecedent is to be understood quantitatively, given that co-reference with a (non-matrix) discourse antecedent is possible, and this seems to hold for bothnull subjects and dative experiencersin Spanish adjunct infinitives. An interesting issue for future research would be to compare quantitatively the co-reference patterns of null subjects and dative experiencers in nonfinite and subjunctive clauses introduced by para and other prepositions, which is beyond the scope of this paper.
In the next section, I shortly discuss some cross-linguistic patterns and propose some tentative solutions, as well as possible ways for future research.

Issues for future research: A note on cross-linguistic patterns
According to the approach outlined above, null subjects and dative experiencers compete for NOC in peripheral adjunct infinitives, which do not require OC. In NOC, non-anaphoric (but non-pronominal) AGR does not obligatorily bind the subject to a matrix antecedent, making available the option of direct C-linking. In this section, I discuss some differences between Spanish and German/Italian in this context. Rather than offering a full-fledged analysis, the aim is to indicate some ways for future research.
In blocking OC dative experiencers, Spanish is different from Icelandic, but similar to German (cf. Masullo 1993, Gutiérrez-Bravo 2006, and patterns with several languages (such as Hindi-Urdu, Marathi, Maithili) that have been described by means of the following generalization from Davison (2008): (43) The Dative Restriction (Davison 2008: 34): In contexts of obligatory control, the embedded verb may not assign its (null) subject dative case.
In the approach to Spanish outlined in the present paper, control of dative experiencers is not categorically ruled out, but the derivation crashes because the theme subject cannot value its features.
However, the question why the possibility of nonfinite psych verbs with locally identified dative experiencers exists in NOC adjuncts in Spanish, but not in a language like German, remains open: (44) Hans i sang den ganzen Tag, ohne ihm *i/x zu gefallen. Hans sang the whole day without him.DAT to like/please.INF '#Hans i sang the whole day long, without PRO i liking it. / 'Hans i sang the whole day, without pleasing him x .' Sentence (44), with the dative experiencer ihm, is impossible with the interpretation of co-reference between the matrix subject Hans and the dative pronoun and only grammatical if Hans controls the null subject. In case of intended coreference between matrix Hans and the dative experiencer, a finite clause with an overt expletive es 'it' would have to be used: (45) Hans i sang den ganzen Tag, ohne dass es ihm i/x gefiel. Hans sang the whole day without that it him.DAT liked/pleased 'Hans sang the whole day long, without liking it.' Italian and German adjunct infinitives, on the other hand, do not have the option of null quasi arguments or postverbal DP subjects. This might indicate that Italian nonfinite T lacks the option of interpretable [iφ:def] (i.e., abstract AGR). Instead, anaphoric AGR is uninterpretable and, therefore, must be checked by a minimal pronoun and is incompatible with a [+R] DP or quasi-argument: (49) … [ CP C LOG/TOP [ TP D [_] / *la Caballé T [uφ:self]/EPP -salire [ vP D [_] v-salire … As I have argued throughout the paper, the impossibility of non-controlled subjects has as a consequence that locally identified dative experiencers are blocked as well.
As a last cross-linguistic point, a reviewer notes that English allows some examples of gerunds with expletive it and a dative experiencer that is co-referent with a matrix antecedent, but English lacks pro: (50) He i added … without it being important to him i .
Interestingly, this configuration involves the preposition without and a gerund. Reuland (1983: 130f) argues for "absolutive P-NP-ing constructions" containing with or without that INFL lacks tense, but it is specified for abstract agreement (i.e., [-T; +AGR]) and, thus, it can assign Case to its subject position: (51) the minister left the pulpit [without [ anything having happened]] (Reuland 1983: 130 [my emphasis]) This would mean that INFL in the configuration (50)/(51) has abstract AGR and is a nominative Case assigner, similarly to the configuration in Spanish and differently from Italian and German adjunct infinitives. Differently from Spanish, however, the subject position is occupied by an overt expletive in English, mirroring finite (non-)pro-drop clauses. Independently of the overt or null realization of the subject position, it seems to be the possibility of non-controlled subjects that makes available the possibility of 'local' co-reference between a dative experiencer inside the nonfinite clause and a matrix antecedent.

Conclusions
In this paper, I have examined dative experiencers in Spanish control infinitives. I have argued that these elements share several properties with null subjects in NOC adjunct infinitives: both elements are in fact pronominal elements which show preferred local co-reference with a matrix antecedent. In the approach of the present paper, this is due to Spanish non-finite T having the option of interpretable [iφ:def], apart from anaphoric AGR. These absorb the EPP but, unlike pronominal AGR in finite clauses, they cannot 'identify' a null subject so that identification is fully achieved via logophoric or topic coordinates in C. EPP absorption has as a further consequence that postverbal DP subjects are licensed.
The possibility of non-controlled subjects (i.e., pro and full DPs) in adjunct infinitives makes it possible for dative experiencers of psych verbs to enter the same referential dependency with C as null subjects. Moreover, given that the dative is higher on thematic hierarchies and scales of topicality, it is the preferred option for local identification with a matrix antecedent in several cases in which the theme subject is an inanimate referent or a CP-clause.
The data discussed here provide additional evidence for an Agree-based Theory of Control in which referential dependencies in nonfinite domains are mediated by the extended projection of the verb -T as well as C. Future research will hopefully clarify further the role that the properties of T and C play in the (im-)possibility of obligatorily and non-obligatorily 'controlled' dative experiencers from a crosslinguistic and micro-parametric perspective.