How bilingual verbs are built: evidence from Belizean varieties of contact Spanish

Bilingual Compound Verbs (BCVs) in Spanish-English bilingual speech, as in (1), are made up of the Spanish do-verb (hacer) and an English-origin component which together form a complex predicate.¹

1. (1) Este  pobre  hizo  invest  dinero.

   dem.sg  poor.man  3sg.pret  invest  money

   ‘This poor man invested money.’

It is generally agreed that hacer, ‘to do/make’, is semantically “light” and has the functional role of carrying grammatical meaning of tense, mood, and aspect, while the English-origin component, invest, carries the weight of semantic meaning (Jenkins 2003; Fuller Medina 2010, 2013; González-Vilbazo and López 2011, 2012). While there is growing consensus that BCVs are little vP structures (Fuller Medina 2007, 2010, 2013; Nakajima 2008; González-Vilbazo and López 2011, 2012), the internal structure of BCVs has generally not been well studied, and they have been considered by some to exemplify a separate bilingual grammar (Romaine 1995, Muysken 2000) or a new structure (Vergara Wilson 2013), while others consider the English-origin component to be code-switched (González-Vilbazo and López 2011) or simply borrowed (Toribio 2001). Furthermore, only dynamic verbs tend to be reported in do-verb constructions, and Spanish-English BCVs are no exception. These structures are highly productive in Belizean varieties of Spanish (Fuller Medina 2005, 2015; Balam et al. 2014); it would therefore be unexpected to find the virtual absence of a particular verb class. Given this productivity and the assertion in the language-mixing literature that any content word is “fair game” to be borrowed (Poplack 1993: 277), the observation that BCVs appear almost exclusively with dynamic English verbs is somewhat puzzling.

The proposal

As the discussion will show, appealing to a separate bilingual grammar to account for BCVs is unnecessary (González-Vilbazo and López 2011, MacSwan 2012). I assume instead that the English-origin component is somehow integrated into Spanish – that is, borrowed – and it is this “somehow” that I attempt to specify. Data shows that BCV hacer is much like the Spanish light verb hacer. It is the phonological spellout of v, retains the lexical meaning ‘to do’ along with the corresponding lexical-aspectual property of dynamicity, and it predicates jointly with its complement (Fuller Medina 2007, 2010, 2013; Gonzalez-Vilbazo and López 2011; Vergara Wilson 2013). However, it does less of the functional work, with its main role being to carry the requisite Spanish inflectional morphology and to specify the doing of V. While the English component in BCVs has been analyzed as a nominal, bare verb, or infinitive (González-Vilbazo and López 2011, Vergara Wilson 2013), I argue that in Spanish-English BCVs, it is inserted into the structure as a full verb. Thus, the English component brings in argument structure with v as part of its structure. Consequently, BCVs are split vP structures where the roles of v are divided between two separate heads.

Furthermore, I rely on well-known proposals that verbs are composed of different flavours or types of little v and a root, and that the lexical-aspectual properties of the root and v must be compatible for verb formation (Cuervo 2003; Folli and Harley 2004, 2007; Harley 2009). I suggest that such a compatibility requirement may be a requisite for BCVs as well. Cast within this type of analysis, the observation that statives are rare in Spanish-English BCVs can be better understood as an incompatibility between some residual dynamic property of the light verb (hacer) and a little v _be on stative complements. Thus, I examine BCV constructions using well-known observations in the literature on verb composition and light verbs, with the aim to better understand how English verbs are borrowed and to shed light on the rarity of stative BCVs.

The remainder of the squib is organized as follows: A brief description of the data is presented in section 1.1. The theoretical assumptions are laid out in section 2, followed by section 3, which introduces the proposed analysis of the structure of BCVs. Sections 3.1 and 3.2 analyze the status of the borrowed component and hacer, respectively. A conclusion and summary discussion are found in section 4.

1.1 Data

All examples, unless otherwise indicated, are taken from production data collected via a picture elicitation task with 22 Spanish-English speakers in Belize for whom this structure is productive (Fuller Medina 2005: 135). Examples marked ungrammatical are constructed by the author, but based on an original grammatical example produced by a speaker. These are judged ungrammatical by both a native speaker consultant and the author, also a native speaker.

2. Theoretical assumptions

I adopt the little v hypothesis (Chomsky 2001), where little v is a functional head which introduces events (Cuervo 2003; Folli and Harley 2004, 2007) and determines the category of the root that is selected (Chomsky 2004: 6). That is, if the nearest c-commanding f-morphemes are little v, Aspect, and Tense, then the root is verbal (Harley and Noyer 1999: 2). The resulting structure is represented in 0 below.²

1. (2) [vP [agent][v'[v][VP [V] DP]]]

This functional head v does not generally have phonological form in English but can be spelled out in some cases, such as in light-verb constructions. In other languages, such as Japanese, v has an overt phonological form. Nonetheless, it is assumed to be present irrespective of the language or type of construction. It follows from this that verbs are composed of at least two parts: a functional head – that is, a light verb or verbalizing head – and a lexical root (Marantz 1997). This functional head is also responsible for introducing events, introducing an external argument, and assigning case to objects of the verb (Burzio 1986, Chomsky 1995). In (3), the root invest originates in the VP but moves up to v to merge with the functional head responsible for the agent argument and accusative case on the DP. Alternatively, as will be shown further below, these roles need not be accomplished by one head and may be shared or split between two heads (see Harley 2017 for a recent detailed analysis of split v).

1. (3) [vP[agent][v'[v invest][VP[V < invest > ] DP]]]

2.1 Flavours of v

Little v also introduces different types of events, suggesting that there are different flavours or types of v (Cuervo 2003; Folli and Harley 2004, 2007; Harley 2009). Thus, verbs are formed by one of four flavours of little v – v _do,v _cause,v _become,v _be – which merge with a root, and the resulting predication is dependent on the lexical meaning of the root and the nature of v. Little v _do introduces dynamic agentive events (Mary ran a marathon), and v _cause introduces dynamic events where the specifier is a causer rather than an agent. Unaccusatives are introduced by v _become and, finally, v _be introduces stative eventualities. Along with lexical meaning, the roots themselves also encode aspectual meaning or aktionsart, leading to selectional restrictions on the type of v with which they will form predicates. The lexical-aspectual property of the root must be compatible with that of the functional head v. The root dance, for example, is both dynamic and agentive and would therefore be compatible with do but not with become or be (Cuervo 2003). The type or flavour of v can be determined by examining its complement, its specifier, and the object-taking properties of the predicate.

3. The structure of BCVs

In the following sections I show that hacer and the English component predicate jointly. I also argue that this borrowed component is verbal rather than nominal and that it is not an infinitive as has often been assumed (e.g., Vergara Wilson 2013).

3.1 The English component

The English component in BCVs is generally not a nominal form (i.e., hacer penalize rather than hacer penalty), and if a determiner or numeral is used to modify it, then the BCV is ungrammatical, as shown in (4).³

1. (4) *Nos  hacían  cinco/el  penalize

   to us do3pl.imp five/the penalize

   *They used to five/the penalize us.

Likewise, in (5), un download ‘a download’ is not a predicating nominal and therefore is not predicating jointly with hacer; consequently, hizo un download does not select el archivo, ‘the file.’ This contrasts with (6), where hacer and download predicate jointly and take a direct object (un archivo grande, ‘a large file’). Thus, BCVs function as constituents or complex predicates rather than as two separate lexical verbs.⁴

1. (5) * Juanita  hizo  un  download  el  archivo.

   Juanita  do3sg.pret  det  download  det  file

2. (6) Juanita  hizo  download  un  archivo  grande.

   Juanita  do3sg.pret  download  det  file  large

   ‘Juanita downloaded a large file.’

This verbal form is often taken to be infinitival and, in fact, in English, the infinitive form may appear bare and is often homophonous to finite forms. However, this is not the case for Spanish, which is morphologically richer and has overt verbal morphology. Borrowed verbs are integrated into Spanish either into the –ar class of verbs (e.g., to type, tipear) or via BCV (e.g., to type, hacer type), and we can assume some regularity in the borrowing process where differences in surface forms are a result of how verbal inflectional morphology is expressed. It cannot be the case that in tipear the borrowed item is an infinitive form to which Spanish inflectional morphology is affixed. Similarly, the English-origin component in BCVs is not an infinitive. Before turning to details of the English component, the status of hacer as a light verb is discussed.

3.2 Hacer: spell-out of little v?

The verb hacer is quite versatile in Spanish, having at least three different functions (Solé 1966) : (i) lexical verb (María hizo un pastel, ‘Mary made a cake’), (ii) idiomatic expression, which may have an encyclopedic entry as a constituent (hacerse el tonto, ‘to pretend to be a fool’), and (iii) functional verb, either a causative (María hizo comer el pastel a Juan, ‘Mary made Juan eat the cake’) or a light verb (hacer fumigaciones, ‘do fumigations,’ ‘to fumigate’). In (i), hacer assigns both external and internal arguments, whereas in the latter cases it assigns an external argument. Thus, it is associated with at least one argument, either an agent or a causer. Note also that even in its light-verb capacity where it predicates jointly with its complement, hacer retains the lexical meaning of doing/making something, which makes it compatible with the nominal complement. The basic properties and functions of hacer in Spanish include: (i) a [+dynamic] lexical-aspectual property, (ii) assignment of an external argument (and accusative case to object DPs), and (iii) carrier of verbal morphology.

In BCVs, hacer behaves most like its Spanish light-verb counterpart as it predicates jointly with the English verbal component as seen in (6), above, and (7)–(9), below. Hacer is dependent for predicaton on this component, which in turn only behaves as a full verb due to hacer. Further, the English component is essential to the meaning of the BCV complex since it specifies the event. This is consistent with the behavior of light verbs and the analysis that roots become full verbs as a result of being selected by a light verb: little v (Marantz 1997). Thus, as in typical light-verb constructions, BCV hacer would be the phonological form for little v. However, this does not fully account for the data.

The data in (1), repeated as (7) below, along with (8)–(9), show that while the verb hacer occurs with an English-origin verbal component in each example, changes in transitivity, agentiveness, and argument structure are observed. Recall that in the previous description of monolingual Spanish, hacer is responsible for introducing the external argument and assigning case. Example (7) is agentive, and dinero has accusative case. So far, BCV hacer appears to behave as in monolingual Spanish and is consistent with little v. However, whether or not there is an external argument does not seem to depend on hacer, even though we would expect it to introduce one, as in monolingual Spanish. Examples (6), above, and (7)–(8), below, all have external arguments, in contrast to (9), which does not, even though they all appear with hacer. In (9), Marta is the theme undergoing the action of falling. I consider this argument to be an internal argument originating in object position in (9), even though it is in subject position, because it patterns with complements. The verb drop, and its particle-verb counterpart drop down, equivalent to fall (in standard English) and Spanish caer(se), alternates with transitive drop (‘Marta dropped the glass’, Marta hizo drop el vaso.)⁵. In both the transitive and unaccusative forms, the object has the role of theme, but in the former it receives accusative case and in the latter nominative. In addition, the argument is completely affected by the falling event in the same way that objects are in transitive constructions such as John ate the apple up. Thus, in (9), there is no external argument. At this point, BCV hacer diverges from Spanish hacer. If hacer does not assign the external argument, then it is not assigning case in (6) or in (7). This suggests that it is doing less functional work and is a lighter form of Spanish hacer. The introduction of the external argument and assignment of case is accomplished through other means.

1. (7) Este  pobre  hizo  invest  dinero.

   dem.sg  poor.man  3sg.pret  invest  money

   ‘This poor man invested money.’

2. (8) Está  haciendo  complain.

   be3sg.pres  do3sg.prog  complain

   ‘He's complaining.’

3. (9) Marta  se  hizo  drop down.

   Marta  cl  do3sg.pret  drop down

   ‘Marta fell down.’

Further indications that BCV hacer does not introduce arguments come from elision. If hacer is elided from any of the above sentences, while they would be degraded, the arguments are recoverable, which is consistent with the analysis so far of the English component as verbal. This contrasts with Spanish, where the complement of hacer in light-verb constructions is nominal and elision results in ungrammaticality. If hacer is elided from (10), for example, then the resulting structure in (11) is ungrammatical. This structure is not well-formed as neither the relation between the arguments nor the event can be recovered. The argument rezos could be specifying an event or could be the complement of a transitive verb (Juan heard prayers). Thus, the arguments are not recoverable in the same way as in BCVs, where a structure like (12) is not rejected, even if it is degraded.

1. (10) Juan  hizo  rezos  en  el  velorio.

   ‘He  did  prayers  at  the  wake.’

2. (11) *Juan  rezos  en  el  velorio.

   *‘Juan  prayers  at  the  wake.’

3. (12) ?? Este pobre invest dinero.

   ?? ‘This poor man invest money.’

Together, the observations regarding the unaccusative example in (9) and elision of hacer point to the English verbal component being responsible for argument structure. If it is entering the derivation as a full verb, then consonant with the analysis that verbs are composed of a root and v, its structure must include a v that assigns category, introduces events, introduces external and internal arguments (in (7), hombre and dinero respectively), and assigns accusative case to objects. And, in fact, we see evidence of this v in (4), hacer penalize, where it is spelled out as the verbalizing morpheme -ize. Thus, the work of little v in BCVs appears to be split between two vs, as shown in (13), where Spanish inflectional morphology is not expressed directly on the English component but rather on hacer. Note that unlike simple verbs, where a root merges with little v to form a verb, the English component does not move up to merge with hacer (Fuller Medina 2007, 2010, 2013; González Vilbazo and López 2011). Since Spanish requires inflectional morphology of tense, mood, and aspect to be spelled out on the verb, a host is required. The verb hacer is selected since it is already available as a light verb (i.e., a bleached form) and is akin to an all-purpose verb.

1. (13) [vP[agent][v'[v hacer][v'[v][VP[V invest]]]]]

I propose that this BCV hacer is an even weaker version of Spanish light hacer. In BCVs, it is bleached of the functional properties of argument introducer and case assigner. In addition, in contrast to Spanish light hacer, which takes nominal complements in forming complex predicates, BCV hacer is not limited to nominal complements and can take verbal ones to form complex predicates. Like its monolingual Spanish light-verb counterpart, however, it retains the lexical meaning ‘to do’ along with the corresponding aktionsart of dynamicity. In other words, it may be less contentful than Spanish light hacer, but it is not vacuous. In BCVs, it is the carrier for verbal inflection morphology and means the doing of V. Within this analysis, hacer could only be a dynamic little v. This hacer may be a further grammaticalized form of Spanish light hacer or may be an adaptation of the existing Spanish light verb for bilingual use.

3.2.1 BCVs: English v and little v hacer

As noted earlier, the type or flavour of v can be determined by examining its complement, its specifier, and the object-taking properties of the predicate. The BCV in example (6) denotes a dynamic event, takes a direct object (un archivo), and has an external argument (Juanita) responsible for the downloading of the file. Consequently, English v may be analyzed as v _do or v _cause. Similarly, the English component in (7) is analyzed as having v _do in its structure. Examples (8)–(9) are both dynamic and intransitive, but (9) does not take an object NP and lacks a volitional agent responsible for the falling, as discussed above. The predicate is unaccusative, and v on the English verb can be analyzed as v _become. Note that drop down, once integrated into Spanish via hacer, must now conform to Spanish syntax as evidenced both by the use of se and its position relative to the verb.⁶ In Spanish, se has various functions, one of which is to indicate that the object is totally affected by the action (e.g., Marta se comió la manzana, ‘Marta ate the apple up’). This affectedness is often expressed by a particle in English.⁷ Without se, the reading that the entity is completely affected by the falling is not available, and both the se-less BCV and Spanish monolingual form (*Marta cayó, ‘Marta fell’) are rejected. Regarding the position of se, this is determined by the finiteness of the verb. Since hizo drop down is finite, as determined by the finiteness of hacer, se is preverbal. The preverbal position also confirms the constituent nature of the BCV (Butt 2004).⁸

The English component is formed under the usual process of a root merging with v – v _do,v _cause,v _become, or v _be – before merging with hacer as a complex predicate. The examples shown thus far exemplify the first three types of v, all of which are dynamic; consequently, there is no source of incompatibility with hacer. Recall that hacer has been analyzed so far as retaining the property of dynamicity. But what of stative BCVs? The only prohibition on monolingual verb formation is the compatibility of root and functional head. Roots like dance or run would not combine with v _be, whereas compatible English roots such as love and know would. Yet, prototypical statives in BCVs, such as (14), are neither reported in the literature nor judged to be grammatical.⁹

1. (14) * Juana  hizo  know  la  respuesta.

   Juana  do3sg.pret  know  the  answer

   ‘Juana knew the answer.’

If it is the case, as I argue here, that hacer is dynamic and that for a BCV to be well-formed hacer must be compatible with the English v in much the same way that roots and vs must share lexico-aspectual properties for a verb to be formed, then the rarity of stative BCVs may be due to incompatibility of English v _be and dynamic hacer. Crosslinguistic evidence is suggestive of such an incompatibility since in other language pairs, two different light verbs are used depending on the aktionsart of the borrowed verb (Muysken 2000). In Punjabi-English, for example, a do-verb is used for more dynamic English verbs, and a separate verb, meaning ‘to be’ or ‘to become,’ is used for less dynamic or stative English verbs (Romaine 1995).¹⁰ It suggests that statives may be borrowed via a bilingual light-verb construction if a light verb with a stative property is available.

4. Conclusion

I have presented evidence that Spanish-English BCVs are complex predicates and that they are split vP structures. The roles of v are split between a Spanish little v spelled out by hacer and a v which categorizes an English root as a verb, introduces arguments and assigns accusative case to objects. This English v is most often null but may also be overt, as shown in examples such as hacer penalize. The verb hacer in these constructions has been analyzed as an even lighter form of the Spanish light verb hacer, and this BCV hacer has the primary functions of carrying verbal inflectional morphology and allowing the English component to predicate fully in Spanish. The borrowed component is a full verb and therefore semantically and structurally rich, bringing in lexical content, argument structure, aktionsart, and transitivity. This suggests that speakers have a full understanding of the English verb that is being borrowed and do not reanalyze or incorporate a semantically impoverished version of the verb. Bilingual speakers appear to make use of a linguistic universal; that is, they exploit vP to borrow English verbs. This, in turn, might explain the productivity and prevalence of bilingual compound verbs in numerous language pairs.

I have also proposed an incompatibility between the two vs in the split vP structure as a possible source for the virtual absence of statives in BCVs. If the two vs must be compatible for the BCV to be well-formed, and hacer is dynamic – meaning, the doing of V – then the rarity of statives in these structures is a logical consequence since, structurally, stative verbs would have a stative v and doing is incompatible with being or the holding of a state.

What appears to be an innovative or special bilingual structure is simply a vP structure that current theories of verb structure and formation can account for. Recourse to a third grammar or bilingual syntax is unnecessary, as is the categorization of bilingual compound verbs as “new” structures per se. The innovation in these structures lies in the bilingual nature of their surface form, in the fact that a vP structure may be built from two different languages, and that speakers make use of a lighter version of hacer to form bilingual compound verbs. While we often appeal to theory to explain bilingual data (see MacSwan 2012 for a review of various approaches), as has been done here, this type of data is, in itself, fertile ground for testing theory (see González-Vilbazo and López 2011, 2012) and for elucidating the details of how distinct grammars come together in bilingual and multilingual discourse.