Is there really an aspectual se in Spanish?

1 Introduction

There is a huge literature on reflexive constructions, both cross-linguistically and within specific languages and language families. The reason for this wide interest in reflexives is the functional overload that these elements show in many languages, where the reflexive marker (henceforth RM) can not only appear in semantically reflexive constructions but can also act as a reciprocal marker (see, for instance, the studies presented in Nedjalkov [2007]), a middle marker (see, among others, Kemmer [1993]) or even as a passive or impersonal marker (see Siewierska [1984] for a typological account). For a smaller number of languages, an allegedly aspectual function of the RM has also been described (see Cornips and Hulk [1996] for Heerlen Dutch and French, Campanini and Schäfer [2011] for French and Italian), an analysis that has been very successful for some Spanish constructions over the last three decades or so (Basilico 2010; Campanini and Schäfer 2011; De Miguel and Fernández Lagunilla 2000; McCready and Nishida 2008; Nishida 1994; Rigau 1994; Sánchez López 2002; Sanz and Laka 2002; Zagona 1996). In this paper, I challenge these claims for an aspectual se in Spanish by identifying some empirical problems that arise in connection with it. In order to do so, I will use corpus data of spoken Spanish from different varieties of Peninsular Spanish, as compiled in COSER (Fernández-Ordóñez 2005).^[1]

The paper is structured as follows. In Section 2, I summarize the main characteristics of the analyses that have identified an aspectual se in Spanish. The data is then briefly described in Section 3. Section 4 focuses on some empirical inadequacies of the aspectual analysis regarding: a) requirements on the DO of transitive predicates that take the RM (Section 4.1), b) the aspectual characteristics and semantic implications of some predicates that take the RM (Section 4.2), and c) those predicates that are claimed not to take the RM (Section 4.3); in Section 4.4, I discuss some predicates omitted from previous accounts. The final section then provides a discussion of the findings and argues for an exhaustive approach to the analysis of these uses.

2 Aspectual se

Although a consensus regarding the different functions that the RM might exhibit in Spanish is far from having been reached, some broad functions, such as the reflexive proper,^[2] the reciprocal, the passive and impersonal, and other valency-reducing functions are identified in one way or another in almost all of the many analyses to be found in the literature (Alcina and Blecua 1975; Cartagena 1972; Gómez Torrego 1992; Martín Zorraquino 1979; Otero 2002; Sánchez López 2002; among many others). Similarly, most authors identify a number of uses that do not fit in any of these categories because they are “facultative use[s] and thus the presence of the [reflexive] clitic does not substantially change the properties of the construction: it neither implies a change in the argument structure […] nor has consequences for the interpretation of any argument” (Sánchez López 2002: 108, my translation).^[3] In these constructions, the RM may attach both to transitive and intransitive verbs.

While traditional accounts did not go much further than merely noting these uses (especially with intransitive verbs) and considering the RM an ethical dative, they have received a great deal of attention in the last twenty-five years. Perhaps the most prevalent tendency in modern accounts (especially within the generative framework) is to attribute an aspectual function to the RM (Basilico 2010; Campanini and Schäfer 2011; De Miguel and Fernández Lagunilla 2000; McCready and Nishida 2008; Nishida 1994; Rigau 1994; Sánchez López 2002; Sanz and Laka 2002; Zagona 1996, and to some extent MacDonald 2017), although some authors have argued that the presence of the RM in these constructions is (also) related to semantic features of the verbal arguments, such as the agentivity of the subject (Aaron and Torres Cacoullos 2005; Armstrong 2013; De la Mora 2011; Di Tullio 2012; García Fernández 2015; Maldonado 1989, 1999, 2000; Rivas 2011; Torres Cacoullos and Schwenter 2008). In this paper, I will concentrate on the former, the aspectual function. Many of the studies addressing these uses focus either on intransitive or on transitive verbs but since they typically assume that the RM bears the same function with both types of predicates, a joint treatment of the different proposals is justified, as I will do in this paper.

The attribution of an aspectual function, typically related to perfectivity, telicity and/or the culmination of the event, is based on contrasts such as those in (1) and (2), taken from the literature. Examples (1a–b) below show that reflexive transitive verbs do not admit bare nouns as direct objects (DOs henceforth), and the contrast in (1c–d) is intended to illustrate that even with definite nouns, the DO of a reflexive transitive verb must be a delimited entity. The contrasts in (2) show that only predicative adjectives and adverbial adjuncts describing the state of the subject during the culmination of the event are admitted with reflexive intransitive verbs.

While many authors have proposed that the presence of the RM with these verbs is related to the aspectual characteristics of the predicate, specific accounts vary as to which aspectual characteristics are attributed to the reflexive pronoun or verb. Early accounts had already noted that reflexive verbs such as morirse ‘to die’, venirse ‘to come’, marcharse ‘to leave’, quedarse ‘to stay/remain’, caerse ‘to fall’ and dormirse ‘to fall asleep’ are perfective, as opposed to their non-reflexive counterparts (Fernández Ramírez 1986). It has also been proposed that what enables the presence of the RM is the telicity of the predicate, this from both a generative perspective (Sanz and Laka 2002) and within the framework of Role and Reference Grammar (Kailuweit 2011). For others, the RM marks the culmination of the event (De Miguel and Fernández Lagunilla 2000; Zagona 1996) or the delimitation of the object (Nishida 1994).

All these proposals share the prediction that only some predicates accept the presence of the RM although exactly which ones varies from author to author. There seems to be some agreement that these predicates are typically accomplishments (Sanz and Laka 2002) although some authors also add delimited states to the list (Nishida 1994), while others exclude predicates where the culmination of the event is not followed by a result state (De Miguel and Fernández Lagunilla 2000). Another point of divergence is that different theoretical accounts may attribute adverbial functions to the reflexive pronoun (De Miguel and Fernández Lagunilla 2000; Zagona 1996) or categorize it as a low applicative (Teomiro 2013) or a light verb (Basilico 2010).

While formal accounts of the so-called aspectual se show some differences in the specific details of the analysis, they also share some core characteristics, four of the most relevant for our purposes being the following:

1. Aspectual se is legitimized by the aspectual class of the sentence in which it appears. Different authors identify different specific aspectual restrictions that affect aspectual se, but these are always related to telicity, perfectivity, delimitation or culmination of the event. In this sense, it is commonly agreed that the event marked by the RM must have an endpoint at which the action is complete.

2. These aspectual restrictions have consequences in the formal properties of DOs in reflexive transitive verbs, which cannot be bare plurals or mass nouns (Nishida 1994; Sánchez López 2002; Zagona 1996), see (1b) above.^[4]

3. As noted above, aspectual se cannot appear with all predicates. Attempts to identify the aspectual feature that allows its presence are not only intended to explain which predicates admit the RM, but also which predicates reject it.

4. It is unclear whether semantic restrictions on the role of the subject also play a role in the presence of the RM. For some authors, a subject’s volitionality or control is higher in reflexive intransitive verbs than in their unmarked counterparts, as shown with estar(se) ‘to be’ (Sánchez López 2002), or the subject needs to be an agent in the case of reflexive transitive verbs (Sanz and Laka 2002). For other authors, the subject need not comply with any restrictions and might also be an experiencer (De Miguel and Fernández Lagunilla 2000).^[5] While recent studies analyzing some of these verbs from a variationist approach have defended the relevance of semantic parameters related to the role of the agent, such as subjectivity and counter-expectation, authors defending the aspectual se interpretation consider aspectual restrictions of the predicate more relevant than other possible semantic considerations:

Taking into account all the cases where the clitic is legitimized, we conclude that the only characteristic of the subject that is relevant to the analysis of these constructions is the fact that the clitic, which has person and number features, must agree with it. Regarding a potential thematic or semantic relationship between the subject of the event and the clitic, which has been postulated in some previous proposals, we believe it does not exist, because the semantic and thematic characteristics of the subject are irrelevant to the presence of the clitic (Sanz and Laka 2002: 322, my translation).^[6]

In summary, aspectual accounts of reflexive verbs with no change of valency in Spanish propose that the RM has a clearly-defined aspectual function and that its distribution is subject to clear-cut restrictions depending on the aspectual characteristics of the predicate. As will be shown, these aspectual accounts tend to focus on a small subset of verbs that optionally admit the RM without changing their argument structure or the interpretation of an argument in Spanish, and hence fall under Sánchez López’s (2002) definition, given above. This is probably due to the fact that, from a methodological perspective, aspectual accounts of the RM are generally built on introspective judgements, typically those of the authors themselves.^[7] While introspection has proved to be a useful technique in the description of systematic grammatical processes, I will argue that this is not so for the description of highly lexically-dependent phenomena, such as appears to be the case with the addition of a facultative RM.

Using introspection as a basis for analysis has led to accounts of aspectual se relying on “clear examples” (Itkonen 2005; Willems 2012) of a few salient verbs. Such clear examples make it possible to propose a clear-cut function for the RM. This method illustrates the fact that humans are better at deducing the general from the particular than the converse case (Nisbett and Borgida 1975). Nevertheless, clear examples can be misleading, because “[t]he human mind does not deal well with nonevents” (Kahneman 2011: 200). Accordingly, we tend to jump to conclusions as soon as we have made sense out of the available evidence, becoming blind to the fact that there might be an absence of important evidence. Moreover, the human mind also has trouble assessing the adequacy of small samples (Tversky and Kahneman 1971) and can easily infer an extreme behavior of a population based on very few examples (Nisbett and Borgida 1975). While these cognitive biases are of little importance when addressing a homogeneous population that shows systematic behavior with little variability, they can have a significant effect on the analysis of more heterogeneous situations.

In this paper, I argue that the strong reliance on a small number of examples, most of which show clear-cut contrasts, has resulted in a skewed sample of the data analyzed, and that as a consequence a number of contexts have been overlooked, some of which might indeed be the most frequent ones in actual speech. This is clearly seen in the COSER data with the verb ir(se) ‘to go’, which is commonly cited in works on aspectual se to show that the presence of the RM might affect the aspectual structure of the predicate (De Miguel and Fernández Lagunilla 2000; Sánchez López 2002; see also Section 4.2.1). Examples such as (3) below, in which the presence of the RM correlates with the presence of a source complement and the verb takes the meaning ‘to leave’, are typically adduced to support this hypothesis. Figure 1 shows the relative frequency of ir (with no RM) versus irse (with RM) in the COSER data in six different contexts: with goal phrases with locative adverbs, preposition a ‘to’ and preposition para ‘for’, with no locative phrases, with goal phrases with preposition por ‘through’, and with source complements with preposition de ‘from, out of’. These data 1) confirm that the RM is most frequent in relative terms with source phrases (and must appear if de means ‘out of’, see Section 4.2.1), 2) show that these are the least frequent contexts overall, and 3) that irse is much more frequent in absolute terms when neither a goal complement nor a locative phrase is present, and where the nuance of leaving need not be present (see Section 4.2.1); that is, the corpus data confirm that the contrast in (3) is correct and show that such a contrast represents an extreme behavior found in only a small subset of examples.

Figure 1:

Frequency of the RM with ir by locative complement in the COSER data.

The aim of the current study is to show the limitations of comprehensive accounts of reflexive verbs with no change of valency in Spanish based on introspective judgements and intuitively selected examples, in that they leave large amounts of evidence unaccounted for. When a comprehensive corpus analysis is performed, the distribution of the RM seems not to be determined by any strict syntactic and semantic constraints, but to be the product of a number of analogical processes related to several characteristics of the verb (often not related to its aspect) as proposed in De Benito Moreno (2015, in revision). For reasons of space, I will focus on aspectual accounts since they are the ones that most consistently propose a comprehensive account of these verbs. Because I believe that corpus-based research is crucial in addressing such empirical inadequacies, this paper is meant as a critique of the method and not of the theoretical perspective adopted in such studies; as such, the criticism is equally applicable to other comprehensive accounts of these reflexive verbs that rely on introspection, such as those of Maldonado (1999, 2000, in which a cognitive framework is adopted.

3 Data and methodology

This study, contrary to virtually all those that have thus far argued for an aspectual nature of se, is based on corpus data. In this sense, it follows more recent work that has adopted a variationist perspective in the study of these forms (Aaron 2003, 2004; Aaron and Torres Cacoullos 2005; De la Mora 2011; Merino Hernández 2019; Rivas 2011; Torres Cacoullos and Schwenter 2008). Most of these studies are ‘agnostic’ towards the aspectual interpretation of the RM or in fact lean towards a subjective reading, often related to Maldonado’s (1999, 2000 cognitive proposal. Although I believe that variationist analyses of this kind are very much suited to the study of reflexive verbs with no change of valency since they exhaustively consider the contexts in which these verbs appear, they also show effects of the introspective bias noted above, in that they focus on very few verbs, virtually all of which have been extensively discussed in aspectual accounts of the RM.^[8]

I have analyzed 139 interviews from the COSER corpus (Fernández-Ordóñez 2005). This corpus consists of semi-directed, in-situ interviews of around 64 min with non-mobile, older rural speakers (both male and female) from various villages in Spain. The mean age of the speakers in this ongoing corpus is 73.4 years. The interviews were conducted in all provinces of Peninsular Spain except Catalonia (because interviews had yet to take place at the time); Galicia and Asturias were also omitted since it is known that reflexive pronouns in their regional languages, Galician and Asturian, as well as the Spanish spoken in these areas, are less productive and appear in fewer types of reflexive constructions (Cidrás Escáneo 1991; D’Andrés 1994; De Benito Moreno 2015, among others). The locations of the different villages included are shown in Figure 2.

Figure 2:

Geographical distribution of the COSER interviews used.

The examples were collected by manually extracting all instances of the relevant verbs from the transcriptions. For some extremely frequent verbs, such as ir(se) ‘to go’ or saber(se) ‘to know’, instances were exhaustively collected only in a subset of the total interviews. This subset includes 54 interviews and has at least one interview per province (De Benito Moreno 2015); the subset is referred to as ‘subcorpus E’ henceforth.

Since the COSER corpus collects spoken and substandard language, the spelling of the examples does not always conform to the orthography of Standard Spanish and shows some of the variable pronunciation features of different speakers. When suggesting that an example is unacceptable, I will avoid the use of the asterisk (*) and instead use double question marks (??) since my experience with corpus data advises me against generalizing my own judgements to other varieties, or even to my own.

Since the goal of the paper is not to propose a new analysis of these reflexive verbs (for this see De Benito Moreno 2015, in revision), the quantitative data are only intended to confirm or disprove the existence of the contrasts proposed in the literature. Thus, no inferential statistics is used.

4 Four problems with current accounts of the so-called aspectual se

In this section I discuss some of the empirical inadequacies that the exhaustive analysis of corpus data show when confronted with some of the most important generalizations that aspectual accounts propose. I will discuss the requirements on the delimitation of the DO of reflexive transitive verbs (Section 4.1), the semantic implications of the aspectual function of the RM in the meaning of the verb (Section 4.2), and the rejection of the RM by predicates that do not conform to this aspectual function (Section 4.3). In Section 4.4, I will discuss some of the evidence that these accounts leave out regarding verbs that do admit the RM but which have not been discussed in previous accounts.

4.3 Excluded predicates

An important prediction of aspectual accounts is that the RM should be impossible with some predicates. In this section I will show that dialectal data contradict such predictions by focusing on entrar(se) ‘to go in’, llegar(se) ‘to arrive’, (des)aparecer(se) ‘to (dis)appear’ and vivir(se) ‘to live’.

4.3.1 Entrar(se)

De Miguel and Fernández Lagunilla (2000) note that entrarse ‘to go in, to enter’ is not expected in their account of culminative se because it conveys an accomplishment not followed by a final state. They do, however, acknowledge that entrarse is possible in some varieties and, despite the fact that they explicitly decide to disregard those varieties in their study, they also claim that the meaning of entrarse is closer to meterse ‘to get in’ than to entrar, which would justify the accommodation of entrarse in their account.^[18] The COSER data, due to their diverse dialectal composition, document several examples of entrarse (15 in total), all of which are perfectly replaceable by entrar (see De Benito Moreno 2015). Moreover, in many of them the replacement by meterse would be either marked or inappropriate^[19]:

(23)

a.

Te	entrabas	solterita	y	te	salías
refl.2sg	went.in.2sg	single	and	refl.2sg	went.out
ya	casadita.
already	married
‘You entered [the church] single and left already married.’
(Cáceres, COSER 1012)

b.

Nos	entremos	[…]	en	Francia	a	las
refl.1pl	went.in.1pl		in	France	at	the
diez	de	la	mañana.
ten	of	the	morning
‘We went […] into France at ten in the morning.’
(Vizcaya, COSER 4504)

4.3.2 Llegar(se)

Sanz and Laka (2002) consider the attachment of the RM to llegar ‘to arrive’ to be impossible. This is a direct result of their claim that the RM is an overt marker of accomplishments and that all unaccusative verbs (except for morir ‘to die’ and caer ‘to fall’) encode achievements. Similarly, De Miguel and Fernández Lagunilla (2000) believe that the absence of a final result state prevents llegar from taking the RM, despite the fact that it has a culmination phase. Again, these authors note that llegarse is possible in some varieties and point out that llegarse has a meaning closer to acercarse ‘to get closer’, which would allow for the inclusion of this variant in their account. However, the COSER examples show that this change of meaning does not adequately explain the distribution of the RM with llegar – all 6 examples of llegarse could be replaced by unmarked llegar, but the replacement by acercarse would be marked or inappropriate in most cases (see also De Benito Moreno 2015):

(24)

a.

Entonces	se	llegaba	el	mes	de	agosto.
then	refl.3	arrived.3sg	the	month	of	August
‘And then August arrived.’
(Salamanca, COSER 3610)

b.

Tú	ahora	que	te	llegues	tú	a
you.sg	now	that	refl.2sg	arrive.2sg	you.sg	to
casa,	te	pegas	la	barrica
house	refl.2sg	hit.2sg	the	loaf
‘Now, when you get home, you eat the loaf [of bread].’
(Zaragoza, COSER 4719)

4.3.3 (Des)aparecer(se)

The form aparecerse ‘to appear’ is explicitly rejected by Sanz and Laka (2002) on the same grounds as llegarse – it is an unaccusative verb that encodes an achievement. Aparecerse, however, is the default variant of the verb when it refers to supernatural subjects such as ghosts in Standard Spanish (Seco et al. 1999), where it is still an unaccusative achievement. Moreover, the responses to a visual questionnaire conducted as an additional source of information to the COSER interviews (De Benito Moreno 2015) show that the forms aparecerse and desaparecerse ‘to disappear’ can show the same meaning as their unmarked counterparts, possible in some varieties of Spanish. They alternate with the unmarked form as shown in (25):

(25)

Ha	desaparecío.	Ha	quitao…	S’-ha	desaparecío.
has.3sg	disappeared	have.3sg	taken.away	refl.3-have.3sg	disappeared
‘It [the picture on the screen]’s disappeared. They’ve taken away… It has disappeared.’
(Huelva, QT 012)

4.3.4 Vivir(se)

The attachment of the RM to vivir ‘to live’ is unexpected in several aspectual accounts for it is atelic (Sanz and Laka 2002) and unergative, that is it has no internal argument (Rigau 1994). However, vivirse (with the meaning ‘to live somewhere’) is also found dialectally in Peninsular Spanish:

(26)

Digo:	“Mire”,	[…]	mis	padres,	yo	me	vivía	más
say.1sg	look.imp		my	parents	I	refl.1sg	lived	more
abajo,	pero	yo	aquí	subí.
down	but	I	here	went.up.1sg
‘I say: “Look”, […] my parents, I lived in the lower part of town, but I came up here.’
(Huércanos, La Rioja, COSER 2506)

4.3.5 Summary

In this section, I have shown that aspectual accounts of the RM cannot accommodate some variation facts of the distribution of the RM. Since aspectual se is deemed to be a facultative marker, it is not unexpected that it is more productive in some varieties than in others. However, we would expect such variation to go in a direction compatible with the alleged aspectual function of the RM, otherwise the behavior of the RM in these varieties (which differs from Standard Spanish only in the behavior of the RM with some verbs, but not all of them) would require a different explanation. However, the data show reflexive verbs that do not fit within the aspectual interpretation of the RM.

This discrepancy could be explained if all the exceptions stemmed from a single focus, since it could be argued that they belong to a dialect with a different (and broader) interpretation of the function of the RM. However, we find no such geographical pattern. On the contrary, each of these verbs seem to show a unique spatial distribution. These distributions only sometimes partially overlap (see Figure 3). Accordingly, we would need a new interpretation of the RM in each dialect that could account for their respective “exceptions”.

Figure 3:

Some reflexive intransitive verbs excluded by aspectual accounts.²⁰

5 Discussion and conclusion

In this paper, I have shown that aspectual accounts of the RM based on introspective data describe and predict stricter contrasts between the marked and unmarked variants than as shown by corpus data: 1) the requirement of a delimited DO, one of the strongest contrasts found in the literature, does not apply to all transitive verbs but only to a few, which have received most attention in the literature; 2) the semantic implications attributed to the aspectual meaning of the RM fall short of general adequacy and do not apply to a large part of the data, and 3) several reflexive verbs which have been claimed to be impossible due to the alleged aspectual function of the RM are found in non-standard dialects that do not differ in other crucial aspects from the varieties typically taken into consideration. That is, aspectual analyses of the RM in Spanish fail to account for a large part of the data as also shown by the existence of reflexive facultative uses that have thus far been ignored in the literature. It seems to me that this situation is symptomatic of one of the crucial flaws in most accounts of aspectual se: this category is almost never defined beforehand; rather, it is simply illustrated by means of a few examples. One exception is Sánchez López (2002), whose definition was given in Section 2, and has been followed here. Yet, as described above, aspectual approaches fail to account for all the data that fit this definition.

In addition, I have argued that the fact that large parts of the data have been disregarded and thus unaccounted for has to do with the methods used in the study of facultative uses of the RM. Aspectual accounts typically rely on introspective judgements and carefully constructed examples that seek to maximize the semantic contrast between marked and unmarked forms. As discussed in Section 2, the use of such clear examples sets the perfect scenario for proposing a clear-cut grammatical function of the RM, which in turn has inadvertently contributed to a blindness to data that do not easily fit such accounts.

While in this paper I have focused on aspectual accounts of the RM because they are the most numerous, it should be noted that other accounts of facultative uses of the RM based on introspective judgements show similar problems. For instance, Maldonado (2000) proposes that the RM in transitive verbs indicates the incorporation of the object in the subject dominion, showing maximal object exploitation and highlighting the agent’s involvement. Accordingly, he proposes the contrasts in (28) below, regarding the combination of the RM and the intensifier de una sentada ‘in one sitting’. While the intuitions on which these contrasts are based might be true, a quick search on social media sites show that they are too strong as shown in (29).^[21] This suggests that introspection is not an appropriate method for the study of these uses because it leads to the overrepresentation of some contexts and the overstatement of some contrasts. As Willems (2012: 678) puts it: “[N]ot only does the variability of language use (in different contexts, by different speakers using different registers, etc.) transcend the ‘’competence’’ of individual speakers, corpus-based evidence also shows that discourse is much more varied and multifaceted than might be supposed on the basis of an individual’s linguistic intuition”.

Introspection, it seems, has effectively slowed down our understanding of these uses since progress has been made through a slow drip of new examples. As noted above (Section 4.1.1), it was not until 2011 that examples of reflexive consumption verbs with bare direct objects were considered (and these are sometimes still ignored, see MacDonald [2017]). Moreover, introspective judgments are especially inadvisable in this case, for the presence of the RM is subject to dialectal, social and stylistic variation and thus shows a distribution that is difficult to access through the researcher’s individual intuition. This might explain the discrepancies regarding the grammaticality of the examples sometimes found in the literature.

As described above, a more recent tendency can be found in the literature with a number of variationist accounts that confirm that aspectual analyses based on observations are not applicable to all the contexts of real speech in which the RM is found (Aaron 2003, 2004; Aaron and Torres Cacullos 2005; De la Mora 2011; Merino Hernández 2019; Torres Cacoullos and Schwenter 2008; Rivas 2011). These usage-based approaches show some clear advantages since they include a larger number of contexts and do not rely on the individual judgements of the researcher.^[22] First, they have been able to show that the contrasts proposed in the literature are not as clear-cut as it seems. For instance, Aaron (2003) finds that irse only means ‘to leave’ in 29% of the cases she analyses. Similarly, both Rivas (2011) and De la Mora (2011) find examples of marked consumption verbs with bare plural objects. Second, they have shown that a particular “construction effect” (Torres Cacoullos and Schwenter 2008: 1471) might be present in the distribution of the RM – Torres Cacoullos and Schwenter (2008) find that subirse ‘to go up’ and bajarse ‘to go down’ most often appear in the construction ‘enter/exit a vehicle’. Last and most importantly, variationist studies have shown that the distribution of the RM with these verbs is not determined by a single factor but that it can be affected by a number of linguistic and even extralinguistic factors, such as dialect or register.

However, they are not free of the influence of accounts based on introspection, something which is clearly seen in the fact that they have focused on a small number of verbs, all of these discussed in the literature previously (consumption verbs, a few movement verbs – ir(se), salir(se), subir(se) and bajar(se) – and morir(se)). That is, a large number of verbs that show the optional presence of the RM are also disregarded in these accounts. That this bias is due to the influence of previous literature is quite clear in Aaron’s (2003) work, who explains the focus of her study as follows: “[t]hough the study included a total of eight verbs, two verbs in particular offer insight into the nature of clitic use variation: ir(se) and salir(se), which will be the focus of this paper” (Aaron 2003: 125). The insight they offer seems to be directly related with the ‘energetic se’ reading proposed by Maldonado (1989, 1999, which Aaron wants to empirically check in her work since her sample contains verbs that are more represented in global quantitative terms (there are 253 occurrences of venir(se) ‘to come’, but only 109 of salir(se) ‘to go out’) as well as regarding the number of marked occurrences (41 of quedarse ‘to stay’ and 27 of venirse, but only 19 of salirse).

From a theoretical perspective, these variationist studies are quite heterogeneous in that some try to offer support for only one of the accounts previously proposed in the literature while others try to conbine several of those accounts and look for evidence in favor of the effect of all of them. This is precisely one of the most valuable contributions of these studies since they try to find empirical evidence that supports the different proposals about the function of the RM. Thus, they not only consider proposals of an aspectual se but also Maldonado’s (1989, 1999, 2000) proposals of an energetic se, of the role of counter-expectation and agent’s involvement, and Clements’ (2006) proposal that the RM is affected more generally by the transitivity of the clause. Another important contribution of variationist studies has been to investigate the role of pragmatic factors such as foregrounding, topicality and speaker’s involvement in the distribution of the RM and to find ways of operationalizing such factors.

However, such heterogeneity complicates the task of combining their findings. First, these studies not always take the same factors into account, even when considering similar verbs and theories. One clear example to illustrate this: although Aaron (2003) – for salir(se) ‘to go out’ and ir(se) ‘to go’ – and Torres Cacoullos and Schwenter (2008) –for subir(se) ‘to go up’ and bajar(se) ‘to go down’ – consider clause type in their analyses as a proxy for pragmatic parameters, Aaron and Torres Cacoullos (2005) – for salir(se) ‘to go out’ – do not. Both De la Mora (2011) and Rivas (2011) consider this parameter (both for consumption verbs) but they distinguish different levels – while De la Mora opposes declarative to interrogative clauses, Rivas opposes main to other clauses.

Second, the results of factors that are similar are not always coherent with each other. For instance, while Aaron (2003) finds that main clauses favor significantly the presence of the RM, no such effect is found by Torres Cacoullos and Schwenter (2008). Sometimes the results are even contradictory. A point in case is animacy of the subject with consumption verbs. Rivas (2011) finds that most of his marked examples have human subjects and, thus, excludes this factor as not being part of the variable context. However, De la Mora (2011) does include the factor and finds that non-human subjects significantly favor the presence of the RM – a result that directly contradicts Rivas’s (2011).

Third, because several proposals about the function of the RM are being tested and because they predict different outcomes, almost every possible result can be used as evidence for one of these proposals. Take, for instance, the interpretation of the results regarding the animacy of the subject with consumption verbs that I just mentioned. Both authors include this factor in order to test if the RM is associated with high transitivity. Rivas’s (2011) results support this theory, while De la Mora’s (2011) are in contradiction with it. However, because De la Mora is testing the theory that the RM might mark counter-expectation as well, she integrates her result as supporting evidence for this theory. Methodologically, this is problematic. A similar methodological issue arises through the fact that often multiple factors are investigated in support of a given theory but only significant results are discussed, without considering what the lack of an observable effect in the some of the factors included implies for the predictions of each theory.

Last, general patterns are also hard to find because the results often highlight the fact that the presence of the RM seems to be affected by highly idiosyncratic factors. As said above, Torres Cacoullos and Schwenter (2008) note that the most important factor determining the presence of the RM is a construction effect. Moreover, almost every study that addresses more than one verb finds important differences among them regarding the frequency of the RM – verb lexeme is the most relevant explanatory factor for the presence of the RM in the two studies that analyse consumption verbs (De la Mora 2011; Rivas 2011) and the second most relevant in Aaron’s (2003) study. A comparison of all these variationist studies also show important differences in the frequency of the RM with each individual verb.

In my opinion, variationist studies are the most promising method to provide a good understanding of the facultative uses of the RM because they considerably enlarge the sample of contexts under consideration. But we must also enlarge the sample of verbs that participate in the alternation. Thus, I argue in favor of turning to exhaustively analyzed corpus data for the study of these uses (see De Benito Moreno 2015, in revision). Such an analysis shows that, while the RM seems to be associated with similar meanings when attached to different verbs (García Fernández 2015), these meanings are quite dependent on the semantics of the verb to which it attaches (De Benito Moreno 2015, in revision). This in turn raises the question of how syntactically determined (as opposed to lexically determined) the presence of the RM is. Moreover, it casts doubt on the assumption that a single function underlies the behavior of the RM in these cases, an assumption adopted by most of the authors that have addressed these facultative uses. As I argue elsewhere (De Benito Moreno 2015, in revision), I believe that an account of facultative uses of the RM must take into consideration other functions of the RM (reflexive, reciprocal, anticausative, etc.), which seem to have served as the source of an analogy for the extension of the RM to other verbs where it has no effect on the valency of these verbs. Such an analogy is syntagmatic and paradigmatic (see Fischer 2008 and De Smet 2012) and is not based on a single parameter but on several, from which the most relevant are the semantic class of the verb, the nature of the subject, and the presence of an external cause.