A new resultative construction in Spanish? A reply to Rodríguez Arrizabalaga

1 Introduction

An important typological distinction among languages relates to the expression of directed motion events (cf. Jonas ran into the house). In this respect, Leonard Talmy (1991, 2000 famously proposed that languages – broadly speaking – fall into two types with respect to how directed motion events are expressed. On the one hand, in so-called satellite-framed languages (e.g. English) the manner of motion is typically expressed in the main verb, whereas the path can be expressed via satellites, which primarily encompass particles and verbal affixes. On the other hand, in so-called verb-framed languages (e.g. Spanish), the path is necessarily encoded in the main verb and the manner of motion can only be expressed via adjunct clauses. This difference in the expression of directed motion events is illustrated below, with English (1) as an example of a canonical satellite-framed language and Spanish (2) as an example of a canonical verb-framed language.

Talmy (2000) expanded his original classification in order to account for result states from all types of events, not only directed motion ones. In directed motion events, the result state is taken to be a change of location that results from the traversal of a path, i.e. in (1a) the bottle ends up in the cave after the floating event is over. Outside this domain, result states also refer to changes of state that hold of a participant after the event is over. For instance, in so-called adjectival resultatives (see Green 1972; Beavers 2011; Broccias 2004; Carrier and Randall 1992; Dowty 1979; Embick 2004; Goldberg and Jackendoff 2004; Hoekstra 1988; Rappaport Hovav and Levin 2001; Kratzer 2005; Levin and Rappaport Hovav 1995; Mateu 2012; Nedjalkov 1988; Randall 1983; Simpson 1983; Washio 1997; Wechsler 2005; Wunderlich 1997) the result state relates to undergoing a change of state which results in a modification of some property of a patient, i.e. x becoming clean, flat, etc. (cf. He watered the flowers flat ≈ He caused the flowers to become flat by watering). In English, such result states can also be expressed via satellites – delimiting the event – which generally encompass APs or PPs, and the manner of action, in this case, is encoded in the main verb, consistently with Talmy’s claim that English constitutes a canonical case of a satellite-framed language.

Spanish, on the other hand (and Romance languages in general; cf. Talmy 2000; Mateu 2002; Mateu and Rigau 2002, i.a.), constitutes a robust case of a verb-framed language insofar as resultatives of the English type, e.g. those in which the main verb encodes a manner (of action) and the result state is expressed via satellites as in (3), are not possible, cf. (4). As in directed motion events, in Spanish the result state – of a change of state, in this case – must be encoded in the main verb, whereas the manner of action can only be expressed via an adjunct clause (5), consistent with Talmy’s proposal.^[1]

Despite Spanish being considered a robust case of a verb-framed language, as per Talmy (1991, 2000, some authors (cf. Beavers et al. 2010; Croft et al. 2010; Filipovic 2007; Fortis 2010; Iacobini and Masini 2006; Martínez Vázquez 2013, 2014, i.a.) have nonetheless questioned the status of Spanish and other Romance languages in regard to Talmy’s typology. According to this view, Spanish appears to have actual cases of resultative constructions with satellite-like results. If this were the case, it would provide evidence contra Talmy’s typology and the well-established phenomenon that Romance languages in general are canonical instances of verb-framed languages in contrast to Germanic languages. In particular, in light of the patterns above, this would be surprising insofar as it would putatively provide evidence against the broadly accepted claim that Spanish constitutes a canonical case of a verb-framed language. An important contribution in this respect is that by Rodríguez Arrizabalaga (2014) (hereafter, RA). RA claims that Spanish has actual cases of resultative constructions by describing a type of a (new) construction found in Spanish corpora that she calls hasta la muerte ‘to death’ construction.^[2] This alleged resultative construction is illustrated in (6).^[3]

RA claims that this construction is the equivalent of the English resultative construction with the PP to death, as illustrated in (7).

It is crucial to note that RA (implicitly) assumes that Spanish has (genuine) cases of so-called strong resultatives of the type found in satellite-framed languages like English (7), despite constituting a robust instance of a verb-framed language. This is because, following the classification originally laid out by Washio (1997), strong resultatives refer to those resultative constructions where the meaning of the main verb and the meaning of the result phrase denoting the result state are to be conceived as independent of each other, i.e. it is not possible to predict the result state from the meaning of the verb. Strong resultatives contrast with weak resultatives, where the meaning of the verb and that of the result phrase are seen as depending on each other insofar as the result state denoted by the result phrase can be generally predicted by looking at the meaning of the main verb. For instance, whereas hard in (9a) is an outcome that can be predicted on the basis of the meaning of the verb freeze – it is the case that things become hard when they become frozen – the result state of black and blue in (8a) does not necessarily follow from an event of kicking.^[4] Examples of strong resultatives are provided in (8) and weak resultatives in (9).

According to this classification, therefore, if Spanish hasta la muerte-type constructions were true cases of resultative constructions, as RA claims, they would then fall under the strong resultative-type classification as proposed by Washio (1997) since in this type of resultative constructions, the verbs encode manners of action (e.g. golpear ‘beat’, apedrear ‘stone’, torturar ‘torture’, acuchillar ‘stab’ etc.) – as in the English to death-type resultatives (7) – and crucially the meaning of the verb and that of the adjunct clause expressing the result state of death are understood not to be mutually related insofar as death need not follow from an event of beating, stoning or torturing.^[5] More importantly, though, if hasta la muerte-type constructions constituted actual cases of (strong) resultative constructions, it would be rather surprising, insofar as Spanish (and Romance in general) disallows strong resultatives of the type found in English, as previously observed. Compare (10) with the Spanish equivalents of the English strong resultatives in (8) above.

In the present paper, we argue against RA’s claim that Spanish constructions of the hasta la muerte-type are to be analyzed as cases of strong resultatives equivalent to those of the English to death-type. Rather, assuming a constructivist approach to argument/event structure, we argue that in the Spanish hasta la muerte constructions as analyzed by RA, the hasta PP is to be regarded as an element providing a bound to an unbounded predicate in the form of a syntactic adjunct external to the argument structure of the predicate. In contrast, English PPs of the to death-type are shown to contribute to the argument structure of the predicate, by providing the final state to a resultative event of change of state.

The paper is organized as follows. In Section 2, we lay out the theory of argument/event structure we adopt as well as our syntactic approach to Talmy’s typology. In Section 3, we discuss the approach toward resultativity as assumed in RA. We show that such an approach is problematic as it fails to account for some crucial generalizations regarding the expression of resultativity in English and Spanish. We then provide our constructivist analysis and show how it better accounts for the differences holding between Spanish and English. In Section 4, we respond to RA’s claim that Spanish hasta la muerte constructions constitute genuine cases of resultative constructions. Section 5 concludes the paper.

2 A constructivist approach to argument/event structure

In this section, we provide a brief overview of the theory of argument/event structure followed in the present paper. We adopt a theory of grammar known in the literature as the constructivist (or neo-constructionist) approach, which finds its grounds in works such as Hale and Keyser (1993, 2002, Marantz (1997), Mateu (2002), Harley (2005), Borer (2005), Ramchand (2008), Mateu and Acedo-Matellán (2012) and Acedo-Matellán (2016). From Marantz (1997), Hale and Keyser (2002), Mateu (2002), Harley (2005), i.a., we adopt the assumption that argument structure’s relations can be described as arising from a limited set of syntactic configurations. From Mateu (2002), Borer (2005), Mateu and Acedo-Matellán (2012), and Acedo-Matellán (2016), our approach takes the idea that syntactic operations are conducted upon two sets of building blocks: functional heads, which are grammatically transparent elements giving rise to semantic construals, and roots, which are elements carrying a purely encyclopedic, conceptual content and providing a syntactically non-transparent meaning.^[6]

In this approach, two basic syntactic structures defining relations between arguments in a predicate can be identified as relevant for our concerns. First, a root can be merged as the complement of an eventive v head, which is a functional head syntactically defining a verb and semantically associated with an eventive reading. This structure gives rise to an activity event, where the root incorporates into the v head in an unergative predicate.^[7]

(11)

John dances.

After the incorporation of the root, the v head may further be associated with a DP complement: in this case, a transitive predicate arises where the direct object is understood as an incremental theme providing a scale to the event predicated by the verb, and the event therefore takes a creation/consumption reading (see Hale and Keyser 2002; Ramchand 2008; Mateu and Acedo-Matellán 2012).

The DP complement can either identify a cognate object or a hyponymous object: in the former case, a root identical to the one which incorporates into v is present (13), while in the latter case, the relation between the direct object and the root incorporating into v is one of hyponymy (12). In both cases, a derivational relation appears to hold between the action named by the (verbal) root and the DP object.^[8]

The second basic structure relevant for our discussion consists of a v head taking a small clause result complement, along the lines of Hoekstra (1988, 1992. The small clause is intended as a phrasal projection headed by a null functional head (labelled here as Pred) defining a relation between its specifier (i.e. the subject of the small clause, which can be understood as the Figure element in Talmy’s 2000 sense) and its complement (understood as the Ground element in Talmy’s 2000 sense). This structure is semantically interpreted as a resultative event where the subject of the small clause predicate, as a consequence of the event, comes to hold a final state which is specified by the element introduced in the complement of the small clause. In this configuration, a root merging with the structure as the complement of the small clause is assigned a resultative interpretation. From this position, the root may incorporate into the v head, giving rise to a resultative verb.

(14)

John flattens the metal.

It is important to note that, under the present account, events of change of state and change of location are taken to involve the same syntactic configuration in a localistic perspective (see Jackendoff 1983; Mateu 2008; Talmy 2000). Thus, the final state of a change of state event is equivalent to the final location of a change of location event.

(15)

John enters.

The activity structure, the creation/consumption structure and the resultative structure described above involve the incorporation of a root from the complement of the v head into the v head, where the root is categorized as a verb. However, a root may also adjoin to v directly through external merge. In the case of a creation/consumption event, this happens when a DP first merges as the complement of the v head (16). Crucially, the effected (i.e. created) DP object is not subject to a cognate or hyponymous relation with the verbal root. This follows from the root being directly merged with v and thus not establishing a derivational relation with the DP complement.

(16)

John smiles his thanks.

In the case of a resultative event, the resultative complement of the small clause does not incorporate into the v head. Instead, it is lexicalized in situ, while v is given phonological substantiation via the external merge of another root (17).

(17)

John hammers the metal flat.

In both (16) and (17), the root adjoined to v is interpreted as providing the manner whereby the event takes place. That is, in (16) the thanks are created by smiling and in (17) the metal becomes flat by hammering (see Acedo-Matellán and Mateu 2014; Embick 2004; Harley 2005; Mateu 2012; Mateu and Acedo-Matellán 2012; McIntyre 2004).

2.1 A syntactic account of Talmy’s typology

Following Mateu (2002, 2012, Acedo-Matellán and Mateu (2013), among others, we claim that the constructivist approach to argument and event structure sketched out in the previous section allows us to provide a structural account of Talmy’s typology. Namely, the difference between satellite-framed languages and verb-framed languages boils down to the absence, in the latter, of the operation of root adjunction to v displayed in (16) and (17).^[9] As a consequence, verb-framed languages never display predicates where the verb is associated with a manner reading which, if taken out of the construction, appears unrelated to the creation or the change of state/location event specified by its complement. Compare this with the following made-up examples from Spanish, illustrating what a creation/consumption structure and a resultative structure of the satellite-framed type would look like in this language.

(18)

*Juan sonríe su agradecimiento.

Intended: ‘John smiles his thanks.’

(19)

*Juan martilla el metal plano.

Intended: ‘John hammers the metal flat.’

As mentioned in Section 1, there have been some authors adopting semantic approaches and a Construction Grammar approach (à la Goldberg 1995), who questioned the consistency of Talmy’s typology.^[10] According to this view, the division between verb-framed languages and satellite-framed languages is not clear-cut and the class of verb-framed languages is merely defined on the basis of the fact that the expression of a manner co-event in the main verb is simply less frequent statistically in these languages than in satellite-framed languages. For instance, those constructions of Spanish where a manner-of-motion verb is followed by a hasta PP (20), or where an effected object indeed appears – prima facie – as complement of a verb denoting a manner of creation (21), are to be regarded as satellite-framed constructions under this view.

(20)

Juan	nadó	hasta	las	rocas.
Juan	swim.pfv.3sg	until	the	rocks
‘John swam to the rocks.’ (Real-Puigdollers 2013)

(21)

Ella	murmura	su	incredulidad.
she	whisper.pfv.3sg	her	disbelief
‘She whispers her disbelief.’ (Martínez Vázquez 2014)

An important caveat is thus in order before proceeding any further. We contend that a distinction must be drawn between what can be interpreted as manner and result in a broad sense, i.e. according to world knowledge, and what is to be interpreted as manner and result structurally, which is the only relevant level of interpretation for Talmy’s typology. From the former point of view, there is no doubt that Talmy’s typology consists in a loose, probabilistic classification of languages displaying many exceptional behaviors within the class of verb-framed languages. For instance, a sentence like (20) clearly implies a result in the sense that, in Juan’s swimming activity, his path reached the rocks and went no farther. However, we claim that linguistically (i.e. structurally), this sentence does not contain a stranded result (i.e. a result which is lexicalized independently of the verb), which crucially explains in turn why this construction is perfectly possible in a verb-framed language like Spanish. As already discussed in Mateu (2012), the fact that Spanish predicates with manner-of-motion verbs and hasta PPs do not involve a result structurally, can be made clear by looking at their Italian counterparts with fino a. An important generalization following from the constructivist approach that we are assuming is that a non-transitive resultative predicate consisting of a v head and a small clause complement displays unaccusative behavior (Hoekstra 1988). This is due to the undergoer of the resultative event (the Figure in Talmy’s sense) being merged as the specifier of the small clause, which in turn gives raise to unaccusative predicates when an external argument is not realized (Mateu 2002). Unaccusative predicates in Italian select the BE auxiliary, in contrast to unergative predicates which instead select the HAVE auxiliary (see Sorace 2000 for an overview). Crucially, Italian equivalents of the Spanish construction in (20) take the HAVE auxiliary, showing that a small clause result-like predicate is not involved in these structure.

(22)

Gianni	ha / *è	ballato	fino	alla	cucina.
Gianni	have.pfv.3sg / be. pfv.3sg	dance.pst.ptcp	until	to.the	kitchen
‘John danced until the kitchen.’ (based on Mateu 2012)

It should thus come with no surprise that hasta PPs in Spanish do not exhibit selectional criteria with respect to the types of motion verbs they are associated with as the structure of these predicates does not involve a small clause complement, and thus these predicates are not to be regarded as resultatives in a structural sense, that is, in the sense which is relevant to Talmy’s typology.

(23)

Juan	caminó/gateó/bailó	hasta	la	cocina.
Juan	walk/crawl/dance.pfv.3sg	until	the	kitchen.
‘Juan walked/crawled/danced until the kitchen.’

In a similar vein, the example in (21) displays another kind of construction which, according to Martínez Vázquez (2014), challenges the robustness of Talmy’s typology since the verb involved (i.e. murmurar ‘murmur’) is used to convey the manner by which one’s disbelief is expressed. We note that the possibility of examples of the type in (21) in Spanish should not be surprising insofar as the verb involved, belonging to the class of the so-called verba dicendi, beside implying a manner component, also strongly implies the production of an utterance which can be regarded as a hyponym of the entity (the murmur) introduced by the (verbal) root. Other examples of the type in (21) discussed in Martínez Vázquez (2014) can be regarded – prima facie – as more problematic for our approach insofar as there appears to be no direct relation whatsoever between the verb and its effected object. For instance, in (24), the expression of a person’s despair is conveyed by means of waving their arms.^[11]

(24)

Bracear	su	desesperación.
Wave	his/her	despair.

‘Wave one’s despair.’ (adapted from Martínez Vázquez 2014)

However, a crucial thing noted by Martínez Vázquez with respect to examples like the one in (24) is that these examples are found in Spanish only if the pragmatic context makes the creation event plausible, i.e. if the context is such that a connection between the actual event of creation and the manner co-event expressed by the verb can be immediately recovered.^[12] For instance, (24) is found by Martínez Vázquez in the context of a football commentary about a player who waved his arms on the field to convey despair. Although this is only a descriptive note in Martínez Vázquez (2014), we claim that the pragmatic compatibility between the event of creation and the manner co-event required for these constructions in Spanish is to be interpreted as a way to recover the derivational relation that is structurally established between the (verbal) root and the effected object in the syntax of these constructions. Thus, the constructions provided by Martínez Vázquez (2014) are felicitous in Spanish as long as they can be taken to involve a structure where the (verbal) root is first merged as the complement of the v head and subsequently incorporates into it, leaving room for a DP to further merge as the complement of the verb. That is, (21) and (24) share their structure with predicates of the type in (12) (i.e. dance a tango) and not with predicates of the type in (16) (i.e. smile one’s thanks).

Having described the theory of argument/event structure adopted in the present paper and how such a theory can effectively account for the differences in the expression of resultativity between satellite-framed languages and verb-framed languages, we now move to review RA’s approach toward resultativity in English and Spanish and show why our constructivist approach should be prefered in accounting for the phenomena observed by RA.

3 A comparison with Rodríguez Arrizabalaga’s approach to resultatives

In this section, we provide an overview of RA’s approach toward resultativity noting that such an approach fails to capture some crucial phenomena regarding the differences in the expression of resultativity between English and Spanish. We then show that a constructivist approach to argument structure can naturally account for the differences in the expression of resultativity between satellite-framed and verb-framed languages.

RA distinguishes two types of verbs capable of entering resultative constructions depending on the semantic relation established between the verb and its direct object. On the one hand, the transitive/unaccusative type displays an object which is semantically an argument of the verb. With these verbs, the removal of the resultative changes the semantic/aspectual connotations of the predication but it does not result in ungrammaticality.

On the other hand, the intransitive type includes verbs that usually appear in unergative predicates. Therefore, the presence of a direct object with these verbs is strictly related to the presence of the resultative element, whose absence makes the sentence ungrammatical.

In light of these differences between verb types, RA (p. 127) claims that the two types of constructions – those involving verbs of the transitive/unaccusative type and those involving verbs of the intransitive type – display “completely different” syntactico-semantic behavior, despite sharing the same linear order of constituents.

In the constructivist framework that we assume, syntactico-semantic behavior does not depend on verbs but on structures only. As a consequence, the syntactico-semantic behavior arising from a construction is predicted to remain unaffected by the verb heading the predication. Put differently, the fact that a verb of the transitive/unaccusative type is capable of entering constructions where the direct object is present to the exclusion of the resultative element does not entail that a resultative construction with this type of verbs is syntactico-semantically different from a resultative construction displaying a verb of the intransitive type. Rather, it simply shows that the considered verb (or better, root, in the present framework) can appear in different constructions, and the detected change of its semantico-syntactic behavior depends on these constructions. This reasoning follows from Hoekstra (1988, 1992 and McIntyre (2004) and is also adopted in Mateu (2012). Thus, despite the contrast between (25b) and (26b), according to these authors examples such as (25a) and (26a) share the same resultative structure, which consists of a v head selecting a small clause complement.^[13]

One important claim by RA, concerning the types of resultative constructions available in Spanish, may however be considered a possible counterargument. RA observes that while English displays resultative constructions with both types of verbs, i.e. the transitive/unaccusative type and the intransitive type, Spanish resultative constructions are only found with verbs of the transitive/unaccusative type to the exclusion of the intransitive type. RA also notices that adjectival resultatives in Spanish are typically restricted to the culinary field and to the field of chromatic changes, as illustrated below.

In RA, the latter restriction is explained on the basis of a semantico-pragmatic constraint, whereby the only resultatives allowed are those involving events whose final state is “extralinguistically far more important than the process causing this state” (RA p. 135). In this respect, consider what RA notes regarding such resultatives:

[W]ithin the culinary realm, there is no doubt that what really matters is not the preparation of a dish per se, but rather its final result. […] Similarly, when a chromatic change is described, what is really significant is the new colour achieved, not the precise method bringing the new colour about. (RA p. 135)

As for the former restriction – i.e. that only verbs of the transitive/unaccusative type are capable of entering resultative constructions in Spanish – although we agree with RA’s descriptive claim, we note that no satisfactory explanation is provided concerning why resultatives with verbs of the intransitive type are ruled out in this language, the only mentioned motivation simply consisting of some (not clearly specified) “general principles governing clausal organization in English and Spanish” (RA p. 133). Insofar as no formalization or account of these so-called general principles is offered, such an analysis – however descriptively accurate it may be – has no predictive power with regard to the differences in the expression of resultativity between English and Spanish.

3.1 A constructivist approach to Spanish resultatives

In this section, we show that the fact that Spanish only displays resultatives of the transitive/unaccusative type is not only just an apparent counterargument to our claim that syntactico-semantic behavior depends on constructions rather than on verbs, but it also comes as no surprise since this is a major prediction of the constructivist approach we are adopting concerning the types of resultatives available in verb-framed languages in a broad sense. In addition, it is worth pointing out that even if a distinction between transitive/unaccusative verbs and intransitive verbs could potentially be relevant when it comes to analyzing resultatives in verb-framed languages, the class of transitive verbs as such is not restricted enough to provide an adequate account of the possible range of resultatives in such languages because there are some verbs that are incompatible with resultative constructions despite belonging to the transitive type. An example illustrating this point is provided by the verb hammer, which, according to RA’s diagnostics, behaves as a transitive type verb, as shown below. Crucially, this verb is incompatible with resultative constructions in Spanish (cf. Mateu 2012).

(28)

a.	John hammers the metal.
b.	John hammers the metal flat.

(29)

a.

Juan	martilla	el	metal.
Juan	hammer.prs.3sg	the	metal
‘John hammers the metal.’

b.

*Juan	martilla	el	metal	plano.
Juan	hammer.prs.3sg	the	metal	flat.sg.m
Intended: ‘John hammers the metal flat.’

According to RA’s analysis, a sentence like (29b) would be ruled out in Spanish because the nature of the event is not such that its final state is “extralinguistically far more important” (RA p. 135) than the process leading to it. We suggest that this explanation is theoretically weak since it is not clear what constitutes a final state that is extralinguistically more important than the process causing it (insofar as no account or explanation is provided regarding what final states are extralinguistically far more important than others). Instead, following Mateu (2012), we argue that only those verbs that allow for a resultative reading by themselves (e.g. pintar ‘paint’ as in [27b]), i.e. in the absence of a distinct resultative element, can co-appear with a resultative element in Spanish. This is predicted to be possible in a verb-framed language like Spanish because the result is provided by the verb, while the alleged resultative is a modifier of the final state encoded by the verb (cf. Acedo-Matellán et al. to appear). Compare (30) below.

(30)

Juan	pintó	la	pared	de	rosa.
Juan	pintar.pfv.3sg	the	wall	of	pink.
‘Juan painted the wall pink.’

Our constructivist approach thus rules out examples like (29b) on structural grounds insofar as, despite martillar in (29a) being a transitive verb, the structure of (29a) does not involve a resultative small clause. Rather, following Acedo-Matellán (2016), we take the structure of (29a) to involve an underlying unergative configuration whose direct object is to be understood in terms of an adjunct to the vP.^[14]

(31)

John hammers the metal.

A resultative reading in (29b) can only be provided by adjoining √HAMMER with the v head, while a distinct element, as is flat in (28b), specifies the result component in the complement of the small clause. However, while this process is available in satellite-framed languages like English, it is not available in verb-framed ones like Spanish (recall [18] and [19]), as previously discussed in detail, hence the ungrammaticality of (29b).

To conclude, it is also important to note that our analysis is fully compatible, and actually sympathetic, with Boas’s (2003) (in RA p. 128) descriptive observation that resultatives with verbs of the transitive type can sometimes be used “to emphasize an end state that is conventionally implicit in the change of state denoted by the verb”. In fact, we claim that those Spanish constructions analyzed in RA as involving a resultative distinct from the verb involve such cases (cf. [27]). This conclusion is also in line with Masullo and Demonte’s (1999) observation (in RA p. 136) that Spanish resultatives “seem to refer to a final state which is implicit in the verbal meaning, thus excluding the causative relationship that is a distinguishing feature of the English resultative construction”.^[15] However, following Mateu (2012) and in contrast to Boas (2003) and Masullo and Demonte (1999), our claim is not merely descriptive but theoretically grounded: the verb-framed behavior of Spanish allows a resultative construction to be realized only through verbs resulting from the incorporation of the resultative element from the small clause complement into the v head. As such, the semantic effect mentioned by Boas (2003) and Masullo and Demonte (1999) follows as a natural consequence of our account, and the fact that a verb is capable of selecting a direct object to the exclusion of the (alleged) resultative argument is contingent on the phenomenon of resultativity.

4 A new resultative construction in Spanish?

With the theoretical background laid out in the previous sections, we now move to the hasta la muerte construction, which constitutes the main case study in RA. The author presents the Spanish hasta la muerte construction, illustrated below, as a resultative construction produced from a direct calque of the English resultative PP to death (e.g. John beat Tom to death).^[16] By means of corpora searches, the author finds that this calque was adopted during the second half of the 20th century, as constructions with the hasta la muerte resultative begin to appear in documents from the 1970s. When analyzing the types of verbs entering the hasta la muerte construction, RA is surprised to find that, contrary to all other resultatives found in Spanish, hasta la muerte not only appears in association with verbs of the transitive/unaccusative type (32), but with verbs of the intransitive type as well (33). This is illustrated below, by means of original data.^[17]

RA concludes that Spanish constructions with unergative verbs and the hasta la muerte PP behave as the so-called fake non-conventionalized resultatives of the English type (sic in RA, following Boas [2003], to indicate resultative constructions with verbs of the intransitive type where no particular restriction on the type of result is imposed by the verb). Moreover, as RA notices, this type of Spanish resultatives does not require the presence of a fake object, in contrast to their English counterparts, as illustrated by the following contrast (see also [33]).

Such a contrast between English and Spanish remains a descriptive notation in RA insofar as the author simply acknowledges its existence and provides a description of it, but no explanation is given with regards to why it holds. However, within our constructivist approach to argument structure, both the appearance of hasta la muerte giving rise to resultative structures with verbs of the intransitive type and the absence of the reflexive in this construction in Spanish receive a straightforward and unified explanation. In what follows, we lay out the analysis of such putative cases of resultative constructions.

Drawing on Mateu (2012), we argue that there is a crucial structural difference between English resultatives of the to death-type and Spanish intransitives with PPs of the hasta la muerte-type (33): while the former involve a resultative small clause, the latter appear as external to the argument structure of the predicate (cf. Section 2.1).

(36)

Bebieron	hasta	la	muerte.
drink.pfv.3pl	until	the	death.
Lit. ‘They drank until the death.’

Crucially, the fact that Spanish intransitive constructions with hasta la muerte, in contrast to English constructions with to death, do not involve a small clause, also explains why English constructions, but not Spanish ones, require a reflexive (cf. [34a] and [35a]): as the sentential subject in English is interpreted as external argument (being assigned the theta role of Agent), a reflexive has to be inserted in the position surfacing as direct object, which is the position occupied by the internal subject of the small clause (see the structure in [37]). In contrast, as no small clause is involved in Spanish predicates, the absence of the reflexive in these constructions is structurally accounted for.

(37)

They drank themselves to death.

In addition, insofar as hasta la muerte PPs are not to be regarded as resultative elements in the syntactic sense, i.e. in the sense that we assume to be relevant to Talmy’s typology (cf. Section 2.1 and discussion below), we predict no restrictions on the type of structures which can occur with this type of PPs in Spanish.^[19] This is corroborated by data provided by RA herself which involve transitive (38), unaccusative (39) and unergative (40) verbs (under RA’s classification), which are all free to join this construction (examples from RA pp. 145–150).^[20]

‘On October 13th 1999, Luis Patricio strangled and knifed Herrero to death.’

In fact, this can be shown to hold not only for this construction but for other constructions involving the P hasta in general, as discussed in Section 2.1. In particular, this strongly suggests that hasta is an element capable of appearing with any type of verb and it can therefore give the illusion of displaying satellite-framed behavior (see Beavers 2008). For instance, hasta can combine with both manner-of-motion verbs, where a directional reading is (strongly) inferred (41), and directed motion verbs (42) (see Real-Puigdollers 2013), giving the illusion of a directed motion event of the English type (cf. John danced into the room).

This strongly argues in favor of an analysis that treats hasta Ps as a way to establish boundaries in events that are otherwise unbounded by default. In particular, as Beavers (2008: 285–286) argues, when hasta is combined with manner-of-motion verbs as in (41), it simply delimits the space in which the event can take place (e.g. in [41a] it is understood that Juan did not swim farther than the rocks), rather than entailing directed motion (cf. also Real-Puigdollers 2013). Thus, as Beavers (2008: 313) argues, hasta-type Ps should not be regarded as satellites which imply a goal or a path, as in satellite-framed languages, but rather as elements which simply provide a boundary to an eventuality which is otherwise unbounded. Consequently, constructions involving hasta-type Ps in verb-framed languages do not violate Talmy’s typology iff the typology is understood as a distinction in the encoding of results structurally, and they are crucially predicted to be fully available and possible in this type of languages.

Last, we want to address another comment from an anonymous reviewer, which in turn takes us back to the discussion in Section 2.1, regarding the fact that hasta la muerte constructions do entail that the action denoted by the main verb causes the death of the object referent (e.g. Juan golpeó a la víctima hasta la muerte, #pero sobrevivió ‘John beat the victim to death, #but he survived’), which, however, does not appear to follow from our structural approach as the anonymous reviewer points out. Recapping the discussion in Section 2.1, we agree with the reviewer that hasta la muerte constructions do indeed entail that the undergoer of the event denoted by the main predicate dies, which, in turn, involves that a result follows from the event of, say, beating as in Juan golpeó a la víctima hasta la muerte ‘John beat the victim to death’. Yet, as discussed in detail in Section 2.1, we contend that a clear distinction has to be drawn between what is to be interpreted as a result structurally and what can be interpreted as a result in a more general, non-linguistically relevant sense. In the latter sense, a predicate like golpear hasta la muerte is of course resultative in the sense that the person who is beaten eventually dies as a consequence of the beating. However, the idea of result arising from this predicate is the product of a logical inference which transcends the linguistic dimension: the result is not encoded in the structure of the predicate. In our approach, languages encode the result by means of a small clause predicate complement of an eventive v head in syntax. Crucially, this structural notion of result is the only one that plays a role when dealing with Talmy’s typology, as it is the only one where the effects of Talmy’s typology actually arise. To this extent, it should not be surprising that Talmy’s typology appears as a probabilistic generalization if a generic, descriptive notion of result is assumed. In our view, however, this fails to capture the crucial fact that Talmy’s typology plays a role in the mind of the speaker, that is, it defines the ways in which a result (in the relevant sense) can be expressed in linguistic predicates by the speaker. Non trivially, once a structural account of Talmy’s typology is assumed, those naturally occurring examples in verb-framed languages that might appear to question the validity of Talmy’s typology are either explained by the absence of a (linguistically relevant) notion of result with respect to the alleged resultative element (cf. hasta PPs) or by the capability of coercing an interpretation of the verb as arising from a process of root incorporation, both in the case of creation/consumption predicates (recall [21] and [24]) and in the case of resultative predicates. With respect to the latter, for instance, a construction like barrer las hojas (lit. ‘to sweep the leaves’, intended as ‘to remove the leaves by sweeping’) is taken to be possible in Spanish insofar as the root √BARR- can be coerced into a resultative reading of removal (thus, in structural terms, insofar as the root can be interpreted as incorporating into v from the complement of the small clause predicate; cf. Mateu [2017]). That is, any possible manner entailment which the root may carry with it as part of its encyclopedic content should not be regarded as playing a linguistically relevant role with respect to Talmy’s typology.

5 Conclusion

In this paper, we have responded to RA’s claim that Spanish has seen the birth of a new resultative construction. After introducing our constructivist view of resultative constructions and the syntactic account of Talmy’s typology deriving from it, we have shown that RA’s Construction Grammar’s account of Spanish resultative constructions is not accurate in that it is not able to correctly predict what verbs are allowed to participate in these constructions. Namely, RA’s claim is that only verbs of the transitive/unaccusative type are typically able to enter resultative constructions in Spanish. However, we have shown that the class of transitive/unaccusative verbs as such is not restricted enough to describe such a behavior. Rather, what is relevant for resultative constructions in verb-framed languages like Spanish is that the result is incorporated into the v head. Importantly, this property has been shown to be independent of transitivity by itself. Subsequently, we have shown that the hasta la muerte construction of Spanish and the to death construction of English involve two distinct syntactic configurations and that only the English construction constitutes a genuine case of a resultative construction, contra RA. We have proposed that the hasta la muerte PP of the Spanish construction is merged as an adjunct external to the argument structure of the predicate and provides a bound to the eventuality introduced by the predicate it merges with. By doing so, following Aske (1989), Beavers (2008), Mateu (2012) among others, we have argued that hasta-type constructions, more generally, do not constitute counterexamples to Talmy’s typology. Rather this type of structures is expected to be available and productive in verb-framed languages insofar as it does not involve the syntactic operation of root adjunction to v which is taken to be relevant to the typology.