Mixed Verbs in Contact Spanish: Patterns of Use among Emergent and Dynamic Bi/Multilinguals

Abstract

The present study provides a quantitative analysis of mixed verbs in the naturalistic speech of 20 Northern Belize bi/multilinguals of two different age groups (ages 14–20 and ages 21–40). I examined the relative frequency of Spanish/English mixed verbs vis-à-vis syntactic verb type and phrasal verbs in mixed verbs. Results showed that the token frequency of mixed verbs was a predictive measure of the relative frequency of ‘hacer + V’ in code-switched speech. In relation to syntactic verb type, it was found that the least productivity in terms of argument structures was attested among the youngest group of emergent bi/multilinguals. For the incorporation of phrasal verbs in mixed verbs, no marked differences were attested in the relative frequency of phrasal verbs across emergent and dynamic bi/multilinguals, but differences did emerge in the semantic nature of phrasal verbs. Findings highlight the fundamental role that adult code-switchers with higher levels of bi/multilingual proficiency play in the creation and propagation of morphosyntactic innovations.

1. Introduction

In the Spanish-speaking world, mixed verbs1 or bilingual compound verbs have been documented in several bi/multilingual contexts involving Creole [1]; Germanic (English: [1,2,3,4,5]; German: [6]); Mayan (Chontal Maya: [7]; Yucatec Maya: [8]); Oto-Manguean (Popoloca: [9], cited in Muysken [10] (p. 193)); and Polynesian [11] languages. In some Spanish contact situations, the light verb bearing morphosyntactic information (e.g., person, tense/aspect/mood features) in these hybrid constructions originates from another language, as in (1), where the inflected Pipil light verb chiw co-occurs with a Spanish lexical verb.

1.	Yah	mu-chiw-ki	arrepentir 2
	PRON.3SG	REFL-do-PST	regret-INF
	‘He regretted (it).’ [12] (p. 144), cited in Wohlgemuth [13] (p. 108)

In most bilingual contexts involving Spanish, however, it is the fully inflected light verb hacer ‘do’ that co-occurs with a donor language lexical verb, as in (2) from Spanish/English code-switching (henceforth CS) 3.

In the present study, I examine the use of ‘hacer + V’ among bi/multilinguals from Northern Belize, a Central American/Caribbean/Latin American context where contact between Northern Belizean Spanish (NBS), English and Belizean Kriol has increasingly become more intense in the last 50 years (for sociohistorical discussion of Spanish in Belize, see Balam [1,2]).

2.	Nunca	he	hecho	witness	un	girls’ fight
	Never	have-1SG	do-PASTPART	witness-INF	a	girls’ fight
	‘I have never witnessed a girls’ fight.’ [3] (p. 254)

The closest equivalent to ‘hacer + V’ in monolingual Spanish is the causative hacer-construction, as in (3). Whereas there is a causer (i.e., ‘he’) and a causee (i.e., ‘me’) in (3), this is not the case in (1) and (2), where there is no element of causation and where hacer only serves as a carrier of morphosyntactic information.

3.	Me	hizo	estudiar.
	CL.INDOBJ	do-3SG.PRET	study
	‘He made me study.’ [15] (p. 300)

Thus, unlike other contact situations (e.g., Bengali in Bengali/English CS: [16]; Welsh in Welsh/English CS: [17]), there is no structural template in monolingual Spanish that parallels the periphrastic ‘hacer + V’ in Spanish/English CS [2,18,19] 4. In light of this fact, innovative forms of ‘hacer + V’ in contact Spanish cannot be analyzed as a mere case of an ongoing language-internal change (but see Backus [20] (p. 334), for Turkish/Dutch CS), as these constructions do not spontaneously emerge in monolingual Spanish contexts.

Considered a potential universal property of CS [21], concomitant questions of when and why these hybrid structures are used still remain unresolved. Some researchers have attributed the incorporation of these structures to lexical deficits ([4,18,22]; but see [19]); whereas others have contended that mixed verbs are particularly used to incorporate phrasal verbs [23,24] or concepts that are culturally novel [21]. Mixed verbs have also been found to be employed in certain discourse contexts such as humor and mockery [25].

An issue that is of particular interest to scholars is who employs these constructions in bi/multilingual communities. In previous work, it has been suggested that speakers with higher levels of proficiency in the component languages [26] or speakers who frequently engage in bilingual language practices [5] use mixed verbs. On the other hand, Wichmann and Wohlgemuth [27] contend that lower levels of bilingualism favor the use of mixed verbs. In their cross-linguistic analysis of different verb borrowing strategies 5 across typologically diverse language contact situations, Wichmann and Wohlgemuth propose a loan verb integration hierarchy (light verb strategy < indirect insertion < direct insertion <| paradigm transfer), where degree of bilingualism in the languages in question plays a central role.

Wichmann and Wohlgemuth’s hierarchy predicts that a higher degree of bilingualism (i.e., increasingly intense language contact, and hence, an increase in societal bilingualism and bilingual proficiency among speakers) leads to a strategy further in the right to be likely selected as the default borrowing strategy [13] (p. 286). Therefore, more bilingualism should likely result in the complete integration or ‘direct insertion’ of donor language verbs, as in (4), where ‘reenlist’ is inserted with no morphosyntactic accommodation. In contrast, less bilingualism should favor the use of mixed verbs (i.e., light verb strategy).

4.	Va	a	reenlist
	Go-3SG.FUT	to	reenlist-INF
	‘He is going to reenlist.’ [22] (p. 154)

Thus far, scant research has been conducted to understand the use of mixed verbs among bi/multilinguals of different age groups and/or proficiency levels. My main concern here is not with when and why mixed verbs are used but ‘how’ they are employed, specifically in comparison to unmixed verbs in naturalistic speech. In order to provide an insight into the relative frequency of mixed verbs across individual speakers and different age groups from Northern Belize, I first conducted a quantitative analysis of 7608 verbs (6440 NBS verbs, 761 mixed verbs, and 407 English or Kriol verbs), extracted from 20 sociolinguistic interviews. Subsequently, I examined mixed verbs in relation to syntactic verb type and the incorporation of phrasal verbs in order to further analyze patterns of use across the two speaker groups that differed in terms of proficiency (i.e., emergent versus dynamic).

This paper is organized as follows. In Section 2, I summarize the work that has been carried out on mixed verbs in Belizean Spanish. Section 3 describes the methodology employed in the present study. In Section 4, I present the results. Finally, in Section 5, I provide a discussion of the main findings and offer concluding remarks.

2. Mixed Verbs in Belizean Spanish

To date, there are only a handful of studies that have investigated mixed verbs in Spanish/English CS [2,3,4,5,18,19,22]. Whereas earlier work on ‘hacer + V’ was primarily descriptive in nature (for California Spanish: [22]; for New Mexican/Colorado Spanish: [18]), recent studies provide more detailed quantitative analyses on the use of these hybrid constructions [3,4,5].

As it relates to Belizean Spanish, Fuller Medina [4] investigated whether mixed verbs were governed by constraints based on stativity and verb frequency. Fuller Medina’s analysis of 35 tokens revealed that speakers generally produced mixed verbs whose English lexical verbs belonged to the class of dynamic verbs, whereas only 10% accounted for stative verbs. She also found that all participants produced mixed verbs containing low frequency English verbs, whereas high frequency lexical verbs such as ‘eat’ and ‘drink’ were unattested. Fuller Medina concluded that her findings suggested that the incorporation of mixed verbs was constrained by stativity and frequency of the lexical verb.

Analyses of data from an acceptability judgment task and interviews with 38 native speakers of NBS, however, revealed that these factors do not constrain the use of mixed verbs [3]. The quantitative analysis of 553 mixed verbs revealed that although ‘hacer + V’ occurred with dynamic verbs, it also occurred with different types of lexical verbs including stative and psychological verbs. Moreover, in oral production, high frequency verbs accounted for 35.9% of lexical verbs. Balam and Prada Pérez [23] further showed that ‘hacer + V’ was productive at the lexical and clausal levels, as mixed verbs were not only attested with a variety of lexical verbs and argument structures, but they favored certain clausal contexts (i.e., subordinate nominal and adverbial clauses).

More recently, Balam [2] provided a cross-generational analysis of mixed verbs vis-à-vis syntactic verb type and pronoun type. The analysis of 1750 mixed verbs, produced by 56 Northern Belize bi/multilinguals, showed that whereas the eldest group produced hacer only with intransitive, transitive and ditransitive verbs, there was a drastic increase in the token frequency of mixed verbs among younger consultants. In addition, there was a notable expansion to other syntactic contexts. For pronoun type, the eldest group produced hacer only with accusative, dative and middle voice pronominal forms. Among younger generations, however, there was an increase in the frequency of use of these forms and a subsequent expansion to other pronominal forms. Although this study provided an insight into the evolution of mixed verbs across generations, it focused on bilingual data; hence, leaving unclear how patterns of frequency attested in mixed verbs compare to those in monolingual verbs.

Endeavoring to provide a more detailed insight into the use of these hybrid structures in relation to monolingual verbs, in the present study, I analyze the production of ‘hacer + V’ and Spanish verbs in a subset of 20 Northern Belize bi/multilinguals (10 second generation and 10 third generation). Although it has been previously claimed that ‘hacer + V’ is frequently used in Belize [2,3,32], this is the first study that provides an analysis of its frequency in relation to monolingual NBS verbs, both across individual speakers and different age groups.

3. Materials and Methods

3.1. Research Purpose

The aim of the study was twofold. First, I sought to investigate whether the relative frequency (i.e., proportion of mixed verbs in relation to total verb production) of mixed verbs in the present corpus showed that ‘hacer + V’ is more frequently used by a certain sub-group of speakers. Wichmann and Wohlgemuth’s [27] hypothesis suggests that particularly emergent bi/multilinguals, who have lower levels of bi/multilingual proficiency, should make frequent use of mixed verbs. Thus, the relative frequency of mixed verbs should be higher among emergent bi/multilinguals pre-disposed to the use of these innovations (i.e., Emergent bi/multilingual Group 2) than among dynamic bi/multilinguals. On the contrary, if mixed verbs are primarily used by speakers with higher levels of bi/multilingual proficiency, as Annamalai [26] suggests, then dynamic bi/multilinguals pre-disposed to the use of ‘hacer + V’ (i.e., Dynamic bi/multilingual Group 2) should make the most frequent use of mixed verbs.

Second, I sought to better understand the similarities and/or differences in the relative frequency of ‘hacer + V’ across the speaker groups in terms of syntactic verb type and the incorporation of phrasal verbs. Although these linguistic factors have not been previously proposed as constraining factors in the use of ‘hacer + V’, the present exploratory analysis examines whether differences in bi/multilingual proficiency result in the differential use of mixed verbs vis-à-vis these two factors. In light of previous observations on the use of these constructions in other contact situations [5,10,26,31,33], a general prediction would be that as a result of higher levels of bi/multilingual proficiency, dynamic bi/multilinguals should demonstrate more frequent, productive, and innovative use of mixed verbs, both in terms of syntactic verb type and phrasal verb forms.

3.2. Participants

The present analysis is based on the data of a sub-set of 20 consultants from a sample of 62 native Northern Belizean bi/multilinguals, whose production of mixed verbs is fully analyzed in Balam [2]. Given that Balam found that ‘hacer + V’ was especially productive among the second and third generations (between ages 14–40), data from a representative group of these two generations were extracted for a more detailed analysis. The sub-set comprises bi/multilingual Maya/Mestizos 6 who were all proficient in NBS, English and Belizean Kriol.

Speakers for the present analysis were selected via purposeful sampling. Selection was based on the number of mixed verbs (i.e., token frequency) produced by consultants, as I sought to examine whether the relative frequency of these hybrid structures in naturalistic discourse was indeed reflective of the differential use of ‘hacer + V’ among speakers. There is the possibility that although adult speakers generally produced more mixed verbs, this token frequency may simply reflect an overall higher production of verbs and not necessarily more frequent use of ‘hacer + V’ (in comparison to younger speakers’ production of mixed verbs). Thus, it could be that although bi/multilinguals from Northern Belize variably produce ‘hacer + V’, when mixed verb production is analyzed in relation to overall verb production, no marked quantitative difference bears out across individual speakers.

In light of this, I examined mixed and unmixed verbs from the five speakers who produced the highest token frequency of ‘hacer + V’ and the five speakers who produced the lowest token frequency of ‘hacer + V’ in both the second and third generations (for details on how generations were determined, see Balam [2] (p. 90)). This yielded four sub-groups, two per generation (see

Table 1. Consultants’ reported patterns of language use and proficiencies.

	Language Use				Self-rated Proficiency
	Bzn Spanish	Eng	Bzn Kriol	Spn/Eng CS	Bzn Spanish	Eng	Bzn Kriol	Spn/Eng CS
	Mean	Mean	Mean	Mean	Mean	Mean	Mean	Mean
EB Group 1 (n = 5)	6.6	5.0	3.0	6.2	6.0	4.8	2.8	4.2
EB Group 2 (n = 5)	6.4	5.6	5.6	6.0	6.6	5.8	6.2	6.4
DB Group 1 (n = 5)	6.2	5.8	5.4	6.2	6.0	6.2	4.8	5.4
DB Group 2 (n = 5)	6.6	6.2	5.8	6.2	5.2	5.4	5.2	5.6

* EB = emergent bi/multilingual, DB = dynamic bi/multilingual, Bzn = Belizean, Eng = English, Spn = Spanish.

). For each generation, the sub-group of consultants who produced the lowest token frequency of mixed verbs comprised the speakers less predisposed to the use of ‘hacer + V’ (i.e., Group 1), whereas those with the highest token frequency were the speakers more predisposed to the use of ‘hacer + V’ (i.e., Group 2).

I labelled the youngest group (between ages 14–20) as emergent bi/multilinguals. This group comprised only full-time, high school students. Following García [34] and colleagues, I use the term emergent bi/multilingual to refer to younger speakers who are still in the ongoing process of acquiring or learning a second or third language in a school setting. Note that although English is the official language of Belize, it is generally learned as a second language via formal schooling [2,35,36,37]. Thus, it is primarily within classroom contexts that students gain higher levels of bi/multilingual proficiency.

On the contrary, I use the term dynamic bi/multilingual [38,39] to refer to adult speakers (between ages 21–40) who have more fully developed, to varying degrees, a wide range of bi/multilingual language practices. It must be highlighted that whereas both emergent and dynamic bi/multilinguals may employ innovative CS practices (see Section 4), it is the latter group that evinces more dynamic use of these practices due to their higher levels of proficiency. In the present sample, second generation consultants had all obtained an associate (n = 6) or bachelor degree (n = 4) and held full-time jobs. I do not use the terms fluent or balanced bilingual, as these terms suggest the notion that bilinguals are equally proficient in their languages, an idea that promulgates a narrow, monoglossic view of bilingualism.

As Table 1 illustrates, across the emergent and dynamic bi/multilingual sub-groups, speakers reported greater use of NBS and Spanish/English CS than monolingual English and/or monolingual Kriol. This pattern is consistent with previous work on Northern Belize, which reveals that Maya/Mestizos have a more positive predisposition toward bi/multilingual language practices than monolingual varieties. A notable characteristic that individualizes the Northern Belize context is that whereas younger speakers generally have positive attitudes toward CS and associate this practice with their identity, the use of monolingual Spanish and/or monolingual English is marked, particularly among adolescents (for details, see [2,40]).

In terms of self-rated proficiencies, there were some differences among the sub-groups. Among the youngest consultants (EB Group 1), for instance, the mean proficiency rating in Belizean Kriol was markedly lower than for the other sub-group of emergent bi/multilinguals. This more than likely is an indication of the changes in these speakers’ linguistic repertoire within the high school context, as it is precisely in their high school years that many Maya/Mestizo adolescents from Northern Belize become proficient and/or fluent in Kriol [2,35,36] and also start engaging in the more frequent use of Spanish/English CS due to increasing levels of bi/multilingual proficiency (see Table 1).

Among dynamic bi/multilinguals, there was greater variation in speakers’ self-rated proficiencies. Notably, speakers had a tendency to rate their Spanish proficiency lower than their English proficiency, a pattern that may be indicative of speakers’ greater confidence or linguistic security when speaking English. Given that English is the official language Belize (albeit not the lingua franca), it is most likely that dynamic bi/multilinguals would more frequently use Standard English for work-related purposes as opposed to younger speakers who would typically use Standard English only in the classroom setting [36,37]. This pattern, therefore, may reflect bi/multilinguals’ increasing use of English in professional domains. It is important to note, however, that dynamic bi/multilinguals do rate their proficiency in Spanish/English CS higher than their proficiency in Kriol, which reveals that bilingual language practices continue to constitute an important aspect of the linguistic repertoire of Northern Belize dynamic bi/multilinguals even as they become fully immersed in the workforce.

3.3. Data

All sentential contexts containing a mixed (i.e., hacer + V) or monolingual verb (i.e., NBS or English/Kriol) were extracted from the 20 sociolinguistic interviews (lasting between 20 and 30 min) and orthographically transcribed. Each mixed or monolingual verb was then individually coded for speaker type (according to the four sub-groups) and syntactic verb type (see categories in

Table 2. Syntactic verb type categories.

Category	Example
Transitive	Abrió la puerta ‘She opened the door.’
Ditransitive	Ella le escribió ‘She wrote a letter to him.’
Intransitive	Ya no volvió ‘He didn’t come back.’
Copulative	Se volvió más dura ‘She became less caring.’
Reverse psychological	Solo perro me gusta ‘I just like dogs.’
Control	Comenzaron a pelear ‘They started to fight.’
Passive	Se enseña dos veces ‘It is taught twice.’

). Mixed verbs were also coded for their co-occurrence with phrasal verbs. Original extraction of the monolingual Spanish verb data set included all ser ‘to be’ (e.g., no es strict ‘He is not strict’) and estar ‘to be’ (e.g., Estaban ella y mi hermano ‘She and my brother were there’) tokens. Exclusion of verbs that are not in variation with ‘hacer + V’ in the corpus was subsequently carried out.

Stative ‘be’ verbs were excluded from the analysis as these verbs are helping verbs that cannot co-occur with the light verb hacer ‘do’. Cases where ‘be’ verbs were attested in the English/Kriol data set (e.g., Overall, it was a good movie) were also excluded. In addition, all modal verbs in the Spanish monolingual (e.g., poder ‘can’, tengo que/de ‘have to’, debo que/de ‘should’, etc.) and the English/Kriol data sets (e.g., can, could, should, etc.) were excluded from the analysis given that ‘hacer + V’ is not in variation with modals 7.

All cases of haber ‘have’, when used in place of ‘there is/are’, were also excluded from the Spanish data set (e.g., Había una piedra grande ‘There was a big stone’), as haber does not occur in variation with ‘hacer + V’ when used as a copula. On the other hand, all cases in the Spanish data set with tener ‘to have’ (e.g., No tiene nada pa’ decir ‘He doesn’t have anything to say’) were included in the analysis, as in previous work [3], tener has been attested with ‘hacer + V’ (e.g.,…porque están haciendo have fun ‘…because they are having fun’) 8. Thus, have-constructions are in variation with ‘hacer + V’. Have-constructions in the English data set (e.g., They have different priorities) were also included for the overall distributional analysis. In the Spanish data set, all established verb borrowings (n = 12) that were partially or fully integrated into the -ar Spanish verb class were included in the analysis (e.g., ponchar ‘to punch (somebody)’, chequear ‘to check/grade school work’), as these verbs have been fully nativized in NBS. Not considering the exclusions of ser (n = 640), estar (n = 715), haber (n = 9) and modal verbs (n = 331), a total of 6440 monolingual Spanish verbs and 761 mixed verbs were subsequently coded for further quantitative analysis.

Categories for syntactic verb type were gleaned from previous studies [2,5,18,19,23]. Coding was carried out using the categories in Table 2. Given that I was not interested in further examining passive mixed verbs, I did not distinguish between the different types of Spanish passives (i.e., stative, eventive, and pseudo-passives).

The examples I provide in Table 2 only illustrate verbs that are embedded in stretches of monolingual NBS discourse. In the corpus, however, Spanish monolingual verbs (e.g., veo ‘I see’, hablando ‘speaking about’, etc.) commonly occurred in stretches of dense, intra-sentential CS, as in (5).

5.

Yo veo que más hacen capture lo que toy tratando de hacer 9 bring across porque lo puedo ver en sus grades, y um, they can quickly hacer recall lo que yo estuve hablando talvez last week, because solo les doy un lee 10 hint…remember hicimos esto. ‘Oh yes!’ y se recuerdan de lo que uno estaba hablando because they were involved.

“I see that they can better learn what I am trying to teach them, because I can see it in their grades, and um, they can quickly recall what I was talking about last week, because you just give them a little hint…remember we did this. ‘Oh yes!’ and they remember what we were talking about because they were involved.” (DB41, male)

While the effect of language mode (monolingual versus switched) in the stretch of speech or intonational unit (IU) immediately preceding and/or following each verb is a phenomenon worth examining (for an analysis of ‘hacer + V’ in relation to IUs, see Wilson Vergara and Dumont [5]), this is not an issue I explore in the present paper. Here I primarily investigate ‘hacer + V’ in relation to overall verb production and different verb types. Backus [31] (p. 216) rightly observes that “[o]riginally the ‘do-construction’ did not receive a great deal of attention in the literature on CS, since the Spanish-English data on which much of it was based, did not include it to a great extent”. The case of ‘hacer + V’ in Northern Belize, however, offers fertile ground to examine this CS phenomenon, as this is a distinctive characteristic of Spanish/English CS in this context [1,2,3].

In the following section, I present the results from the quantitative analysis that was carried out.

4. Results

4.1. Relative Frequency of Mixed Verbs

I sought to determine whether the use of ‘hacer + V’ was most frequent in a certain sub-group of speakers. Recall that Wichmann and Wohlgemuth’s [27] loan verb integration hierarchy suggests that particularly in cases of less bilingualism should the light verb strategy be employed. In line with this contention, it was expected that the relative frequency of ‘hacer + V’ would be higher among emergent bi/multilinguals (particularly among those predisposed to the use of mixed verbs) than among older, more proficient bi/multilinguals.

The distributional analysis showed that there was variable use of mixed verbs (see

Table 3. Verb production among emergent bi/multilinguals.

	Spanish		Mixed Verbs		English/Kriol		Total
	N	%	N	%	N	%	N
Group 1
EB01	194	97.0	6	3.0	0	0.0	200
EB02	372	97.9	8	2.1	0	0.0	380
EB03	381	97.2	11	2.8	0	0.0	392
EB04	314	95.7	13	4.0	1	0.3	328
EB05	454	95.8	20	4.2	0	0.0	474
Group 2
EB06	376	91.7	34	8.3	0	0.0	410
EB07	344	90.1	38	9.9	0	0.0	382
EB08	298	78.0	58	15.2	26	6.8	382
EB09	250	77.9	64	19.9	7	2.2	321
EB10	426	76.5	67	12.0	64	11.5	557

* EB = emergent bi/multilingual, DB = dynamic bi/multilingual.

and

Table 4. Verb production among dynamic bi/multilinguals.

	Spanish		Mixed Verbs		English/Kriol		Total
	N	%	N	%	N	%	N
Group 1
DB01	269	94.7	15	5.3	0	0.0	284
DB02	205	88.7	19	8.2	7	3.0	231
DB03	159	70.7	19	8.4	47	20.9	225
DB04	347	91.1	19	5.0	15	3.9	381
DB05	363	81.4	22	4.9	61	13.7	446
Group 2
DB06	343	68.3	46	9.2	113	22.5	502
DB07	204	70.6	51	17.6	34	11.8	289
DB08	362	84.8	61	14.3	4	0.9	427
DB09	466	83.8	66	11.9	24	4.3	556
DB10	313	71.0	124	28.1	4	0.9	441

* EB = emergent bi/multilingual, DB = dynamic bi/multilingual.

). In both the emergent and dynamic bi/multilingual groups, there were speakers who less frequently produced ‘hacer + V’ and those who more frequently produced ‘hacer + V’. Importantly, however, it was not the case that emergent bi/multilinguals who were more predisposed to ‘hacer + V’ (i.e., Group 2) produced mixed verbs the most. Furthermore, the relative frequencies of mixed verbs in the youngest sub-group (i.e., Emergent bi/multilingual Group 1) were the lowest in the sample, ranging from 2.1%–4.2% (see Table 3); hence, indicating that adolescents are actually the least likely to employ these structures in naturalistic speech.

The data also revealed that there were cases where the token frequency of mixed verbs did not exactly match the relative frequency of ‘hacer + V’. For example, although EB01 produced the least number of mixed verbs, the relative frequency of ‘hacer + V’ in this speaker’s discourse was actually higher than that of EB02 and EB03 (3.0 versus 2.1 and 2.8 respectively). Similarly, in the case of EB10, although this speaker produced the most mixed verb tokens, his production of mixed verbs in relation to his overall verb production was actually lower than that of EB08 and EB09 (12.0 versus 15.2 and 19.9 respectively).

A similar phenomenon was attested in the dynamic group (see Table 4). Although DB09 produced more mixed verbs than DB07, DB09’s production of ‘hacer + V’ in relation to overall verb production was actually lower than that of DB07 (11.9 versus 17.6 respectively). These inconsistencies reveal, therefore, that the token frequency of ‘hacer + V’ may not necessarily be perfectly reflective of subtle interspeaker differences in the relative frequency of ‘hacer + V’.

Crucially, the relative frequency of mixed verbs does capture the general distinction between the two sub-groups that were identified (Group 1 = speakers who are less predisposed to mixed verb use; Group 2 = speakers who are more predisposed to mixed verb use) in the emergent and dynamic bi/multilingual groups.

There were other notable patterns. For instance, in the emergent bi/multilingual Group 1, none of the speakers’ relative frequency of ‘hacer + V’ was higher than 5%, whereas in the dynamic bi/multilingual Group 1, none of the speakers’ relative frequency of mixed verbs was less than 4.9%. This shows that even though both of these groups are less predisposed to the use of mixed verbs, there are quantitative differences between them. On the other hand, among speakers who were predisposed to mixed verbs in the emergent and dynamic bi/multilingual groups, there was an overlap in the relative frequencies of mixed verbs, with the former sub-group ranging between 8% and 20% and the latter ranging between 9%–28%. In terms of the relative frequency of ‘hacer + V’, therefore, speakers from Group 2 in both the emergent and dynamic bi/multilingual groups are more comparable than Group 1 of the two respective groups.

Also notable was that emergent bi/multilinguals made very limited use of English/Kriol verbs in English/Kriol utterances, a pattern which reflects both speakers’ emergent bi/multilingualism and the marked status of Standard English [2,37,44]. Among dynamic bi/multilinguals, the use of English/Kriol verbs was higher than among their emergent counterparts, but Spanish verbs clearly remain the most frequent, consonant with reported patterns of language use (see data for dynamic bi/multilinguals in Table 1).

The fact that the token frequency of ‘hacer + V’ generally matched its relative frequency across emergent and dynamic bi/multilinguals suggests that the token frequency of mixed verbs can be taken as a predictive measure of overall verb switching. Results also showed that it is not the case that all or most emergent bi/multilinguals primarily use monolingual Spanish verbs nor is it the case that some dynamic bi/multilinguals use more mixed verbs than monolingual Spanish ones. In both groups, there was variable use of mixed versus monolingual Spanish verbs, but overall, Spanish verbs were overwhelmingly used. None of the speakers’ production of mixed verbs in relation to overall verb production was 50% or more. In fact, even in the data from the most prolific user of ‘hacer + V’ (see Table 4), mixed verbs only represented 28% of total verb production. This percentage confirms that even among the most pervasive users of ‘hacer + V’ in this context with favorable social conditions for dense CS practices, Spanish monolingual verbs still constitute the majority of verbs employed in naturalistic discourse.

4.2. Syntactic Verb Type

In order to further analyze how consultants used ‘hacer + V’ in naturalistic speech, mixed verbs were subsequently examined vis-à-vis syntactic verb type and the incorporation of phrasal verbs. Given this study’s focus on Spanish/English CS and the marked use of English in the corpus, I focus here on the comparison between the monolingual Spanish and mixed verb data sets. Results are presented in

Table 5. Syntactic verb type of verbs produced by emergent bi/multilinguals.

		Group 1		Group 2		Total
		N	%	N	%	N	%
Transitive	mixed	29	4	122	14	151	9
	monolingual	767	96	750	86	1517	91
		796		872		1668
Ditransitive	mixed	4	3	13	8	17	6
	monolingual	144	97	140	92	284	94
		148		153		301
Intransitive	mixed	24	4	113	17	137	11
	monolingual	598	96	560	83	1158	89
		622		673		1295
Copulative	mixed	0	0	3	4	3	3
	monolingual	41	100	66	96	107	97
		41		69		110
Reverse psych.	mixed	1	2	4	9	5	5
	monolingual	60	98	40	91	100	95
		61		44		105
Control	mixed	0	0	5	4	5	2
	monolingual	103	100	133	96	236	98
		103		138		241
Passive	mixed	0	0	1	17	1	13
	monolingual	2	100	5	83	7	87
		2		6		8

and

Table 6. Syntactic verb type of verbs produced by dynamic bi/multilinguals.

		Group 1		Group 2		Total
		N	%	N	%	N	%
Transitive	mixed	54	7	183	19	237	14
	monolingual	693	93	799	81	1492	86
		747		882		1729
Ditransitive	mixed	6	6	29	16	35	12
	monolingual	101	94	149	84	250	88
		107		178		285
Intransitive	mixed	31	7	116	18	147	13
	monolingual	412	93	541	82	953	87
		443		657		1100
Copulative	mixed	1	4	5	17	6	11
	monolingual	23	96	24	83	47	89
		24		29		53
Reverse psych.	mixed	1	3	3	7	4	6
	monolingual	29	97	38	93	67	94
		30		41		71
Control	mixed	0	0	4	3	4	2
	monolingual	78	100	127	97	205	98
		78		131		209
Passive	mixed	1	12	8	44	9	35
	monolingual	7	88	10	56	17	65
		8		18		26

.

Table 5 below shows that among emergent bi/multilinguals, ‘hacer + V’ was fully productive with different syntactic verb types only among speakers who were more predisposed to the use of mixed verbs (i.e., Group 2). These consultants produced ‘hacer + V’ with all seven types of argument structures, but particularly with transitive (14%), intransitive (17%) and passive verbs (17%). In contrast, the youngest speakers did not produce ‘hacer + V’ with copulative, control or passive verbs.

Table 6 reveals a similar trend for dynamic bi/multilinguals. Among speakers who were more predisposed to the use of mixed verbs, ‘hacer + V’ was attested with all seven argument structures, but particularly with intransitive (18%), transitive (19%), and passive verbs (44%). Notably, in contrast to their Group 1 counterparts in the emergent group, dynamic bi/multilinguals who were less predisposed to the use of ‘hacer + V’ did produce mixed verbs with all syntactic verb types, except control verbs.

Passive mixed verbs such as the stative passive in (6) and the pseudo-passive in (7) favored the incorporation of ‘hacer + V’. In contrast, control verbs like comenzar ‘to start’ and aprender ‘to learn’, as in (8) and (9), favored ‘hacer + V’ the least. Thus, verb switching was more frequent in the complement infinitive verb context, as in (8) and (9), than in the preceding control verb (e.g., que tienes que hace strive to balance… ‘that you have to strive to balance…’).

6.	Mi	class	‘taba	hecho,	like,	overrun	by girls … solo había	five boys
	My	class	be-3SG	do-PASTPART,	like,	overrun-PASTPART	by girls … only were	five boys
	‘My class was overrun by girls … there were only five boys.’

7.	Los	accounts	se	hacen	handle	diferente
	The	accounts	CL.PAS	do-3PL.PRES	handle-INF	differently
	‘The accounts are handled differently.’

8.	Y	comienzan	a	hacer	quarrel	con	el	teacher
	And	start-3PL.PRES	to	do-INF	quarrel-INF	with	the	teacher
	‘And they start quarreling with the teacher.’

9.	Tienes	que	aprende	a	hacer	open	accounts
	Have	to-2SG.PRES	learn	to	do-INF	open-INF	accounts
	‘You have to learn how to open accounts.’

Two points are relevant here. First, passives were very infrequent in the Northern Belize corpus, in line with the trend for monolingual Spanish in general [45], but they were nonetheless a favorable syntactic context for mixed verbs. Passive mixed verbs were primarily produced by the dynamic bi/multilingual who was the most prolific user of ‘hacer + V’ (i.e., DC10), whereas none of the youngest consultants produced a passive mixed verb. This suggests that the most innovative mixed verbs are used by code-switchers with higher levels of bi/multilingual proficiency. Second, control verbs were more frequent than passives, yet they were not a favorable context for ‘hacer + V’ across speaker groups. Despite their considerably frequent production in the corpus, these verbs primarily occurred as Spanish monolingual verbs (see Table 5 and Table 6), suggestive of the fact that control verbs are the syntactic verb type most resistant to morphosyntactic innovation in contact Spanish.

Overall, the data revealed that in terms of syntactic verb type, ‘hacer + V’ was found to be highly productive among all speakers with the exception of the youngest consultants; hence, suggesting that adolescents are the least likely to employ ‘hacer + V’ in innovative structures such as passive mixed verbs.

In the following section, I analyze the data in relation to phrasal verbs in mixed verbs.

4.3. Phrasal Verbs in Mixed Verbs

I further examined the incorporation of phrasal verbs 11 in mixed verbs. Given that Balam and Prada Pérez [23] found that bilexical multisyllabic verbs (i.e., phrasal verbs) favored the use of ‘hacer + V’, a question that arises is whether speakers who more frequently produce mixed verbs also make more frequent use of phrasal verbs in mixed verbs. In previous work, White [46] highlights that phrasal verbs present difficulty to second language learners of English (also see [47,48]), in particular because of their unpredictability, non-compositionality, polysemy, and non-universality. Crucially, among native speakers of English, phrasal verbs are very productive [48] (p. 66); thus, the coining of novel phrasal verbs occurs easily [49]. This raises important questions about the notion of linguistic creativity and the use of phrasal verbs among bi/multilingual code-switchers.

Table 7. Emergent bi/multilinguals’ production of phrasal verbs in mixed verbs.

	N of Mixed Verbs with Phrasal Verbs	Total N of Mixed Verbs	% Containing Phrasal Verbs	Total N of Verbs	% Containing Phrasal Verbs
Group 1	4	58	7	1774	0.2
Group 2	20	261	8	2052	1.0
Total	24	319	8	3826	0.6

and

Table 8. Dynamic bi/multilinguals’ production of phrasal verbs in mixed verbs.

	N of Mixed Verbs with Phrasal Verbs	Total N of Mixed Verbs	% Containing Phrasal Verbs	Total N of Verbs	% Containing Phrasal Verbs
Group 1	5	94	5	1567	0.3
Group 2	24	348	7	2215	1.1
Total	29	442	7	3782	0.8

below present the distribution of phrasal verbs in mixed verbs across speaker groups. Table 7 reveals that there is a seeming increase in the use of mixed verbs containing phrasal verbs among emergent bi/multilinguals more predisposed to ‘hacer + V’. Importantly, however, the frequency of mixed verbs containing phrasal verbs in relation to the total number of mixed verbs and the total number of verbs in the corpus does not markedly differ between the two sub-groups.

Similarly, among dynamic bi/multilinguals, the relative frequency of mixed verbs containing phrasal verbs does not markedly differ between the two sub-groups (see Table 8). Thus, while there were clear inter-group differences in the relative frequencies of mixed verbs within the emergent and dynamic bi/multilingual groups (see Table 3 and Table 4), there were no marked differences in the relative frequencies of mixed verbs containing phrasal verbs. Especially in relation to the total number of mixed verbs, little difference was attested between the two sub-groups in the emergent and dynamic bi/multilingual groups. Only in relation to the total number of verbs was the difference more pronounced.

The lack of marked inter-group differences in both groups suggests that across age groups and proficiency levels, the relative frequencies of phrasal verbs in mixed verbs remain rather stable. Thus, contra our hypothesis, speakers who made more frequent use of mixed verbs did not employ phrasal verbs more frequently when using ‘hacer + V’ in their switched discourse.

The following examples illustrate the use of phrasal verbs alongside the light verb hacer in the speech of emergent (10–12) and dynamic bi/multilinguals (13–15).

10.	Siempre	hacíamos	hang out	after	break
	Always	do-3SG.IMP	hang out-INF	after	break
	‘We always hung out…after the school break.’

11.	El	test	es	que	te	hace	bring down	más
	The	test	be-3SG.PRES	what	CL.INDOJB	do-2SG.PRES	bring down-INF	more
	‘Tests (i.e., low scores in tests) are what bring down your average more.’

12.	Después	hace	find out	que	hizo	destroy	el	whole planet
	Afterwards	do-3SG.PRES	find out-INF	that	do-3SG.PRET	destroy-INF	the	whole planet
	‘But afterwards, he finds out that he destroyed the whole planet.’

13.	…porque	se	hacen	slip off	em veces	pero	los	entiendo
	…because	CL.MID	do-3PL.PRES	slip off-INF	sometimes	but	them	understand-1SG.PRES
	‘…because they lose focus (or control) sometimes, but I understand them.’

14.	…pero	yo	lo	voy a	hace	cut up	to do
	…but	I	it	will-FUT	do-INF	cut up-INF	everything
	‘…but I will cut it (the chicken) up.’

15.	…people	se	‘tan	haciendo	wake up,	‘tan	haciendo
	…people	CL.MID	be-3PL.PRES	do-PROG	wake up-INF,	be-3PL.PRES	do-PROG
	realize,	“Okay,	esto	‘ta	haciendo	go on…”
	realize-INF,	Okay,	this	be-3SG.PRES	do-PROG	go on-INF
	“…people are waking up…they are realizing, ‘Okay, this is going on...’”

Although an in-depth semantic analysis of phrasal verbs in mixed verbs is beyond the purview of this paper, it is noteworthy that speakers with the lowest levels of bi/multilingual proficiency only produced mixed verbs containing phrasal verbs that were school-related, as in (10) and (11). None of the speakers in this sub-group employed ‘hacer + V’ to express abstract ideas. In contrast, speakers who were more predisposed to the use of mixed verbs in the dynamic bi/multilingual group employed ‘hacer + V’ with phrasal verbs that related to their jobs (e.g., write up (a file), branch off, call in (sick), etc.), the home domain, as in (14), and abstract ideas as well, as in (15). This strongly suggests that Northern Belize bi/multilinguals who more frequently use mixed verbs also capitalize on their semantic resources in their linguistic repertoire more than adolescent speakers. The encoding of these nuances in the expression of colloquial speech and abstract ideas adds another layer of complexity and innovation to the code-switched speech of speakers who employ ‘hacer + V’ more frequently.

Noteworthy were salient cases where mixed verbs evinced both simplification and innovation. There were some examples where the verb used seemingly derives from an English phrasal verb but was used in novel ways. For example, in (16), the verb ‘use’ may have likely derived from the English phrasal verb ‘get used to’. In this case, the speaker capitalizes on the lexical and syntactic resources in her linguistic repertoire. While the speaker deletes the particles ‘get’ and ‘to’ from ‘get used to’, she nonetheless builds syntactic/semantic congruence (in the sense of Sebba [43]) between ‘get used to’ and its Spanish equivalent acostumbrarse.

16.	Puedo	hablar	un	poquito	como	ellos	ahora…ya	me	hice	use
	Can	speak	a	little bit	like	them	now…already	CL.INDOBJ	do-1SG.PRET	use-INF
	‘I can speak a little bit like them now…I have gotten used to it.’

The middle voice –se from acostumbrarse is skillfully integrated into the mixed verb. In this regard, we can see that bi/multilinguals establish cross-linguistic semantic parallels between lexical items from their languages even if they are structurally different, and they are able to merge linguistic features from their languages in order to devise novel, hybrid structures (also see (13) and (15)). This phenomenon was particularly attested in the switched discourse of very skilled code-switchers who frequently produced ‘hacer + V’ (for relevant discussion, see Balam [2]).

In the following section, I recapitulate the study’s main findings, and I discuss the implications of these findings.

5. Discussion and Conclusion

In the present study, I analyzed bi/multilinguals’ production of ‘hacer + V’ vis-à-vis the relative frequency of mixed verbs, syntactic verb type and phrasal verbs in mixed verbs. Results revealed that among emergent and dynamic bi/multilinguals, the token frequency of mixed verbs does serve as a predictive indicator of the relative frequency of these hybrid structures in naturalistic speech. Thus, the higher the token frequency of mixed verbs, the more dense and frequent is CS at the verb phrase level among bi/multilinguals. Support was not found for the hypothesis that particularly speakers with lower levels of bi/multilingual proficiency make more frequent use of mixed verbs. In fact, the frequency (i.e., both token and relative) of ‘hacer + V’ among the youngest consultants was the lowest.

In relation to syntactic verb type, speakers who most frequently produced mixed verbs were also the ones who produced ‘hacer + V’ in more syntactic contexts; hence, revealing that the productivity of ‘hacer + V’ in terms of argument structures is not a community-wide phenomenon but one that is associated with certain speakers. Mixed verbs were found to be the least productive in terms of syntactic verb type among the youngest consultants. Thus, whereas adolescents in Northern Belize may accept novel forms of ‘hacer + V’ (for intuitional data, see Balam et al. [3]), they do not spontaneously produce them. In contrast, the other three sub-groups employed ‘hacer + V’ in all syntactic environments, with the exception of control verbs (Dynamic bilingual Group 1), which seem to constitute an unfavorable context for mixed verbs across proficiency levels.

Lastly, in terms of phrasal verbs, the data revealed that there were no marked differences in the relative frequency with which emergent and dynamic bi/multilinguals used phrasal verbs in mixed verbs. However, unlike the youngest speakers who only employed phrasal verbs that were school-related, more proficient bi/multilinguals employed phrasal verbs when speaking about a variety of domains (e.g., jobs, home, school, etc.).

In the ensuing sections, I discuss the implications that these findings have on our current understanding of (i) the frequency of mixed verbs in CS varieties, and (ii) the role of bilingual proficiency in the creation and propagation of linguistic innovations among child versus adult code-switchers.

5.1. Frequency of Mixed Verbs in CS Varieties

In the present corpus, none of the speakers’ mean proportions of mixed verbs constituted 50% or more of total verb production (see Table 3 and Table 4). The distribution of mixed verbs ranged from 6–124 tokens and percentagewise in relation to total verb production from 2.1%–28.1%. Even though the token frequency of mixed verbs among Northern Belize consultants seemed notably high in comparison to speakers from other contact situations (e.g., [19,31]), its frequency in relation to Spanish verbs in the present corpus was consistently low.

In previous work, Myers-Scotton and Jake [51] point out that switching a verb may carry a higher production cost than switching a noun given that verbs play multiple grammatical roles in the construction of bilingual clauses. Unlike nouns, for instance, verbs assign thematic roles. According to Myers-Scotton and Jake, the grammatical roles that verbs play may explain why in several bilingual corpora, nouns have been found to be the most frequently switched elements, whereas verbs are much less frequently switched. In support of Myers-Scotton and Jake’s view, data from Northern Belize show that even among bi/multilinguals who engage in dense CS practices, verbs are not as frequently switched as nouns.

On the contrary, nouns can be so frequently switched that mixed determiner phrases can be greater than the corresponding number of monolingual Spanish determiner phrases employed in naturalistic discourse (for quantitative analysis of mixed determiner phrases in Northern Belize CS, see Balam in press [52]). This illustrates not only the very frequent switching of nouns in Northern Belize bi/multilingual discourse, but it highlights the clear difference in terms of the pervasiveness with which nouns and verbs are switched. Data from Northern Belize suggest that whereas mixed determiner phrases can be quantitatively unmarked12 in CS communities [52], mixed verbs typically remain marked (i.e., they constitute the minority of tokens in relation to overall verb production) even in cases of intense language contact.

It is important to note that even though the use of bi/multilingual language practices in Northern Belize has drastically increased in the past decades due to more intense language contact and higher levels of proficiency in English and Kriol [2,35,36,52], only the use of mixed verbs has increased but not the use of other strategies to borrow/switch verbs. For instance, the use of indirect insertions (e.g., chequear ‘to check/grade school work’) has remained very limited (for similar results in Spanish/English CS in New Mexico, see Wilson Vergara and Dumont [5]). This supports the observation that across bilingual corpora, mixed verbs are more frequently used than direct verb insertions/switches [51] (p.11). At the same time, however, mixed verbs remain constrained in the frequency with which they are used in naturalistic CS.

One factor that may favor the cross-generational use of mixed verbs is precisely the hybrid structure of these constructions. Mixed verbs may seem to require more “integrational effort” [13], but this strategy is actually more parsimonious [3] and productive than other verb borrowing/switching strategies. Importantly, ‘hacer + V’ is not subjected to the same phonological restrictions that indirect insertions are subjected to. To illustrate, consider Sobin’s [53] study on verb borrowings in Texas Spanish. In the 41 established borrowings he identifies, more than 68% are derived from English words ending either in the voiceless alveolar stop /t/ (e.g., trostyar ‘to trust’), the voiceless velar stop /k/ (e.g., kukyar ‘to cook’) or the voiceless bilabial plosive /p/ (e.g., waypyar ‘to wipe’), whereas no loanblend is derived from an English verb ending in a vowel or glide. Thus, whereas English verbs ending with voiceless plosives and other consonants are integrated into Spanish, words ending with vowels, diphthongs, and glides are seemingly blocked from being integrated into the Spanish –ar verb class (e.g., hizo agree ‘he/she agreed’ versus *agreear/agreeió; hizo identify ‘he/she identified’ versus *identifayar/identifaió). Since mixed verbs are not morphophonologically integrated into Spanish, they are not subjected to the same phonological limitations. In this regard, ‘hacer + V’ facilitates the rapid and effortless switching attested among skilled code-switchers.

In addition, mixed verbs may be a universally preferred CS strategy in verb phrases because they are malleable to further innovation and optimization of available syntactic, semantic and lexical resources. As the Northern Belize data have shown, mixed verbs offer the kind of working ground that allows code-switchers to exploit linguistic creativity; in contrast, other strategies (e.g., direct insertions) are more limited in this regard due to their inherent lack of hybrid morphological structure.

5.2. Bi/multilingual Proficiency and Morphosyntactic Innovation

Findings from this study strongly suggest that speakers’ level of bi/multilingual proficiency is an important factor in the use of mixed verbs and the emergence of linguistic innovation. It must be highlighted that Wichmann and Wohlgemuth’s [27] view that lower levels of bilingualism result in the use of mixed verbs are supported by child bilingual data but not post-adolescent and adult data. Crutchley’s [54] recent analysis of narratives from 14 Panjabi/English bilingual children showed that mixed verbs were especially attested among ‘beginner bilinguals’ with limited bilingual competence. These beginner bilinguals generally produced utterances that were less grammatically complex than those attested among more proficient bilingual children who primarily produced monolingual Panjabi utterances and less mixed verbs. In Crutchley’s [54] (p. 20) view, the type of CS among beginner bilinguals “exploit[ed] the simplest structures on offer in both languages”, but it was not associated with grammatical complexity and/or morphosyntactic innovation.

The situation with adolescent, post-adolescent and adult bi/multilinguals presents a different scenario. Elsewhere I have argued that particularly high levels of bi/multilingual proficiency catalyze the use and evolution of mixed verbs (for relevant discussion, see Balam [2]). The present data confirm that the most simplistic use of mixed verbs was attested among adolescent, freshman or sophomore high school students (i.e., emergent bilinguals with the lowest levels of bi/multilingual proficiency). In terms of ‘hacer + V’ production, this was evidenced in the fact that the relative frequency of mixed verbs was the lowest among adolescents. At the syntactic level, this was evidenced in adolescents’ non-usage of mixed verbs with passive, control or copulative verbs. At the lexico-semantic level, this was evinced in their incorporation of phrasal verbs in mixed verbs that specifically pertained only to the school domain.

In previous work, the use of phrasal verbs has been found to be associated with proficiency [56] and/or fluency [57] in English. For instance, in her study on the production of phrasal (transparent and non-transparent) verbs by 29 native Spanish-speaking adult ESL students of three different levels of proficiency, Gaston [56] (p.20) found that an increase in English proficiency corresponded with “an increased understanding of the syntax and semantics” of phrasal verb usage. It is not surprising, therefore, that among more proficient bi/multilinguals, we find not only a higher level of syntactic complexity in speakers’ discourse (as evidenced through argument structures), but we also find the incorporation of phrasal verbs that relate to a wider variety of semantic domains. Thus, an increase in bi/multilingual proficiency seems to be accompanied by corresponding changes in the semantic nature rather than the frequency of mixed verbs containing phrasal verbs.

The cross-generational data in Balam [2] and the present study’s results show that ‘hacer + V’ has developed from less complex to more complex syntactic and pronominal contexts (i.e., from intransitive and transitives to passives and control verbs; from accusative clitics to other pronominal forms). We can posit, therefore, that higher rather than lower levels of bi/multilingual proficiency have contributed to this evolution [2]. To this, we can also add that the more frequent use of ‘hacer + V’ is specifically associated with frequent CS. In their study, Wilson Vergara and Dumont [5] found that speakers who produced the most bilingual IUs were also the speakers who particularly employed ‘hacer + V’, suggesting that skilled code-switchers who frequently engage in CS are precisely those who employ ‘hacer + V’ the most.

We can surmise, therefore, that speakers with higher levels of bi/multilingual proficiency are the ones who employ ‘hacer + V’ in the most innovative ways. In the case of Northern Belize, only speakers with higher levels of bi/multilingual proficiency employed passive mixed verbs, which were unattested among the eldest [2] and youngest consultants. Furthermore, these speakers capitalized on the semantic productivity and polysemy of phrasal verbs (for relevant discussion, see Balam and Prada Pérez [23]); hence, revealing the creative manipulation of the structure and meaning of English phrasal verbs and idiomatic expressions.

Doǧruöz and Backus [42] (p. 58) underscore that in the process of copying (in the sense of Johanson [58], this refers to the importation of elements from another language), speakers’ perception of semantic equivalence plays a key role, “[particularly] where speakers perceive a transparent link between translation equivalents.” From examples such as (13) and (15), it is clear that speakers with higher levels of bi/multilingual competence engage in this search for semantic equivalence, which then results in the production of innovative, hybrid structures that combine different elements from their languages. Thus, the exploitation of similarities between language varieties, which Gardner-Chloros and Edwards [59] (p. 108) argue takes place in CS, particularly occurs among post-adolescent and adult bi/multilinguals who frequently engage in CS.

In previous work some scholars have associated linguistic creativity and innovation specifically with children or adolescents. For example, in their Minimalist analysis of mixed verbs in German/Spanish CS, González-Vilbazo and López [6] assert that the use of structures such as ‘hacer + V’ reveals that speakers are not necessarily restricted to the available features in the component grammars. They argue that this kind of linguistic creativity is evident in cases such as Creole genesis and the formation of sign languages, where features which are absent from the input languages child speakers are exposed to emerge. In their view, “children resort to ingredients provided by Universal Grammar that are absent from the input” [6] (p. 846). González-Vilbazo and López suggest that novel forms of ‘hacer + V’ emerge among bi/multilingual children. Moreover, in previous sociolinguistic research, adolescents have been described as the ‘movers and shakers’ in linguistic change [60,61]. Eckert [61] views adolescents as key players in the process of language change and phonological innovation.

Data from Northern Belize CS, however, show that precisely because of the fundamental import of bi/multilingual proficiency in the use of mixed verbs, children and adolescents may actually be limited in their ability to engage in the kind of “conceptual work” Sebba [43] refers to when he notes that code-switchers are able to actively build congruence or switch junctures in CS (for relevant discussion, see Edwards and Gardner-Chloros [21]). In previous work, Balam [2] (p. 95) found that ‘double hacer’, the most innovative yet infrequent type of mixed verb (e.g., No he hecho learn hacer cook nada ‘I have not learnt how to cook anything’; …que no haga choose hacer study business ‘…that s/he doesn’t choose to study business’), was only attested among highly proficient post-adolescent and adult speakers. If we consider the fact that adolescents in the present sample did not use innovative forms of mixed verbs (passive, copulative and control structures), then more than likely ‘double hacer’ did not emerge among them. Instead, these “unconventional constructions” (in the sense of Doǧruöz and Backus [42]), seem to be an early stage of propagation among highly proficient bi/multilinguals who more frequently employ ‘hacer + V’.

We also cannot attribute novel kinds of mixed verbs such as ‘double hacer’ or passive mixed verbs to the eldest generation, as this generation had limited proficiency in English and produced very few mixed verbs in comparison to the younger generations [2]. There is no doubt, therefore, that the more frequent and innovative use of Spanish/English mixed verbs in Northern Belize is a phenomenon that took place within the last four to five decades. If we take CS as reflective of a high degree of bilingual competence [62,63,64,65,66], then it follows that morphosyntactic innovations in this context emerged precisely among highly proficient adult code-switchers who were able to most creatively exploit the available resources in their rich linguistic repertoire (for relevant discussion, see Muysken [67]). The fact that previously unattested forms (e.g., ‘hacer + V’ in passive and control structures) emerged in the absence of structural equivalents in the component languages reveals the element of creolization that characterizes mixed verbs [30]. In generative terms, it shows that adult code-switchers may resort to and creatively use the “ingredients” or principles provided by Universal Grammar (for relevant discussion, see Bhatia [68]).

Crucially, findings from Northern Belize reveal that language change takes place during language use [69], and that adult grammars are flexible rather than fixed. Therefore, “both innovation and propagation of language change…occur in language use by adult speakers” [70] (p. 59). It is through language use among skilled code-switchers who fluidly engage in a wide range of bi/multilingual language practices that morphosyntactic innovations emerge and subsequently conventionalize. Bybee ([69] (p. 535) rightly emphasizes that “there is a general misconception among linguists and others than language change takes place during language acquisition” when in reality, children are not in a social position to have the kind of influence on adults so as to generate language change (for relevant discussion, see Bybee [71]; Croft [70]). To this, we can add that children and adolescents in bi/multilingual communities, while being proficient in their native languages, are simply not in the best position to effect morphosyntactic change as they are still in a nascent stage of dynamic bi/multilingualism. My view is that while children can acquire canonical forms of mixed verbs (i.e., light verb alongside verb infinitive) and use ‘hacer + V’ with more conventionalized argument structures (i.e., transitives and intransitives), they do not use more innovative mixed verbs (e.g., mixed verbs in passive or control structures) until they have a certain level of bi/multilingual proficiency to naturalistically produce these forms.

It is only among highly skilled post-adolescent and adult code-switchers that speakers exploit the most complex structures in their linguistic repertoire, which allows them to create novel switch sites (e.g., mixed verbs in passive and control structures), build structural congruence between verb equivalents (e.g., merging English phrasal verbs and Spanish middle voice as in (13) where ‘slip off’ and descarrilarse yield ‘se hacen slip off’) and encode their speech with semantic nuances (e.g., reciprocity, reflexivity, etc.). Bhatia and Ritchie [63] (p. 20) remind us that “language mixing is indispensable for creativity and is the defining feature of the bilingual mind/brain.” Certainly, this “defining feature” is especially instantiated among post-adolescent and adult code-switchers who frequently engage in dense CS practices in contexts where bi/multilingual language practices are not only unmarked but valued as well.

In sum, although the prevalence and productivity of mixed verbs are licensed at the societal level, they are clearly capitalized on at the idiolectal level. Speakers with higher levels of bi/multilingual proficiency are precisely those who employ more innovative forms of ‘hacer + V’. At the same time, however, it must be underscored that higher levels of bi/multilingual proficiency alone do not translate to the pervasive use of mixed verbs. Older speakers who were less predisposed to the use of mixed verbs (i.e., Dynamic bi/multilingual Group 1) reported a higher proficiency in English than in NBS, yet their production of ‘hacer + V’ was markedly lower in comparison to other adults; revealing that there are dynamic bi/multilinguals in Northern Belize who make limited use of mixed verbs even after they become more proficient in English. On the other hand, there is another group of dynamic bi/multilinguals that makes extensive use of mixed verbs in switched discourse. It is this group which most likely comprises the locus of morphosyntactic innovation and change in Northern Belize.

Undoubtedly, the present study’s findings support the notion that intra-sentential CS can indeed serve as a vehicle of language innovation and change, a view held by several researchers [20,72,73]. In particular, as it relates to mixed verbs, highly proficient post-adolescent and adult code-switchers are the ushers of cross-generational morphosyntactic change. Findings also support the view that the grammar of CS is emergent [67] and flexible in nature, and novel CS structures that cannot be attributed to any of the component languages may develop [30], particularly in contexts where CS is valued as an everyday communicative resource. Syntactic hybridity, therefore, is not an oddity, but a natural outcome of CS contexts where social conditions allow this language practice to flourish across time. This paper makes a call for further research on bi/multilingual contexts where CS is not stigmatized at a socio- linguistic and political level, as we are most apt to learn about the complex and dynamic nature of CS precisely from these contact communities (for relevant discussion, see Bhatia and Ritchie [74]).