Preposition Stranding in Spanish–English Code-Switching

Abstract

This study tests the acceptability of preposition stranding in the intrasentential code-switching of US heritage speakers of Spanish. Because languages vary when extracting determiner phrases from prepositional phrases, known as preposition stranding or p-stranding, a contrast arises for Spanish–English bilinguals. English allows p-stranding, but in Spanish the preposition is traditionally pied-piped with the DP. Heritage speakers of Spanish, though, have shown variability, with child sequential bilinguals requiring said pied-piping, but simultaneous bilinguals allowing p-stranding in Spanish. Participants (n = 24) completed a written acceptability judgment task with a 7-point Likert scale. The task included code-switched sentences (n = 16) with p-stranding, switching from either English to Spanish or vice versa, with comparison monolingual equivalents for Spanish (n = 8) and English (n = 8) included as well. The results found that the simultaneous bilinguals accepted p-stranding in both languages, while also showing no restriction in either code-switching condition. Child sequential bilinguals, however, showed the expected monolingual distinction between Spanish and English, and p-stranding was only accepted with Spanish determiner phrases extracted from an English prepositional phrase (i.e., Spanish-to-English). These findings support the previously reported differentiation between simultaneous and child sequential bilinguals regarding p-stranding, while expanding it to code-switching.

1. Introduction

Languages vary when extracting determiner phrases (DPs) from prepositional phrases (PPs) (Law 2006; Salles 1995). English allows for such extraction, referred to as preposition stranding or p-stranding, as shown in the sentence in (1). We see that the DP what friend originates in sentence-final position as the complement of the preposition with. However, the DP does not remain in situ as it is part of an embedded wh-question, which requires the wh-phrase to move to a higher position within the embedded complementizer phrase (CP). This movement consequently ‘strands’ the preposition at the end of the sentence; the DP has left a trace in this position and has been moved to CP.

Spanish traditionally does not allow for p-stranding. Consider the sentences in (2). We see that, unlike in English, stranding the preposition is generally considered ungrammatical1 (2a). When wh-movement is required, the preposition is pied-piped along with the wh-element, and as such the entire PP moves as one syntactic unit (2b).

(1)	Chad doesn’t know [DP what friend]i Kevin is traveling [PP with ti].

(2)	a.	*	Él	no	sabe	[DP qué	amiga]i	ella	está	viajando	[PP con ti].
			he	not	knows	what	friend	she	is	traveling	with
			‘He doesn’t know what friend she is traveling with.’

	b.	Él	no	sabe	[PP con	[DP qué	amiga]]i	Sergio	está	viajando.
		he	not	knows	with	what	friend	Sergio	is	traveling
		‘He doesn’t know with what friend she is traveling.’

Given this distinction in p-stranding availability2, a contrast arises for Spanish–English bilinguals, and it is unclear how this syntactic difference manifests itself when the two languages are mixed. Yet to be tested experimentally is the availability of p-stranding in intrasentential code-switching (CS). If one language permits such a construction while the other rejects it, what happens syntactically when the two languages are combined in the same utterance? In other words, is it possible to extract a Spanish DP out of an English PP (3a) or vice versa (3b)?

(3)	a.	Fernando3	no	sabe	[DP qué	amiga]i	Kevin is traveling [PP with ti].
		Fernando	not	knows	what	friend
		‘Fernando doesn’t know what friend Kevin is traveling with.’

	b.	Chad doesn’t know [DP what friend]i	Sergio	está	viajando	[PP con ti].
			Sergio	is	traveling	with
		‘Chad doesn’t know what friend Sergio is traveling with.’

The current study provides experimental evidence that p-stranding is available in Spanish–English CS, but it is mitigated by the age of acquisition (AoA) of English of the speaker, as well as the direction of the switch. Specifically, child sequential bilinguals show no restriction whatsoever regarding p-stranding, accepting such a construction in English, Spanish, and CS (in either direction). Simultaneous bilinguals, on the other hand, show an asymmetry between their two languages, accepting p-stranding in English and rejecting it in Spanish; when mixing the two, they only accept it in Spanish-to-English contexts.

2. Background

2.1. P-Stranding in English and Spanish

Wh-elements in English occupy a higher syntactic position than where they are base generated, generally considered to be the specifier of the CP (Chomsky 1986), as shown in (4a). If the wh-element originates as the complement of a PP, it can be extracted, stranding the preposition in situ in its lower position (Law 2006; Salles 1995), as shown in (4b).

(4)

a.

[DP What]i did you buy ti?

b.

[DP What money]i did you buy it [PP with ti]?

In addition to simple wh-movement (4), p-stranding in English can also occur with embedded wh-movement (5), as well as with relative clauses (6).

(5)	I don’t know [DP what friend]i you went shopping [PP with ti].

(6)	Amy is the friend ([DP who]i) I went shopping [PP with ti].

Heavily stigmatized “in the eighteenth century as being colloquial, inelegant, improper, or even harsh, […] ever since, end-placed prepositions have been frowned upon in grammar books and usage guides” (Yáñez-Bouza 2006, p. 1). Nonetheless, p-stranding continues to be a common syntactic feature of English (Quirk et al. 1985). It is not the only option, though, as pied-piping can alternatively be used, as shown in (7), being employed more often in formal discourse (Biber et al. 1999).

(7)

a.

I don’t know [DP what store]i she got it [PP from ti].

b.

I don’t know [PP from [DP what store]]i she got it ti.

For most Spanish speakers, p-stranding is disallowed entirely, requiring the preposition to be pied-piped with the DP (Law 2006). Syntactically speaking, this construction directly mirrors the p-stranding alternative available to English speakers, with the crucial distinction being it is the only option. Pied-piping occurs with Spanish simple wh-movement (8), embedded wh-movement (9), and relative clauses (10).

(8)	a.	*	¿[DP Qué	dinero]	lo	compraste	[PP con ti]?
			What	money	it	bought.2s	with
			‘What money did you buy it with?’

	b.	¿[PP Con	[DP qué	dinero]]	lo	compraste ti?
		With	what	money	it	bought.2s
		‘With what money did you buy it?’

(9)	a.	*	No	sé	[DP qué	amiga]i	fuiste	de	compras	[PP con ti].
			Not	know.1s	what	friend	went.2s	of	purchases	with
			‘I don’t know what friend you went shopping with.’

	b.	No	sé	[PP con	[DP qué	amiga]]i	fuiste	de	compras ti.
		Not	know.1s	with	what	friend	went.2s	of	purchases
		‘I don’t know with what friend you went shopping.’

(10)	a.	*	Irene	es	la	amiga	que	[DP Ø]i	fui	de	compras
			Irene	is	the	friend	that	wh-	went.1s	of	purchases
			[PP con ti].
			with
			‘Irene is the friend (that/who) I went shopping with.’

	b.	Irene	es	la	amiga	[PP con	[DP quien]]i	fui	de	compras ti.
		Irene	is	the	friend	with	whom	went.1s	of	purchases
		‘Irene is the friend with whom I went shopping.’

How do we account for p-stranding availability across languages? Specifically, why is it that English has both options available (i.e., p-stranding and pied-piping), whereas Spanish is limited to just the one (i.e., pied-piping)? Law (2006) proposes that the availability is “related to the independent grammatical property of [the determiner] incorporating into [the preposition]” (p. 633). He argues that languages like Spanish (including other Romance languages and Germanic languages except English) are subject to a syntax-morphology-interface condition where “elements that undergo suppletive rules must form a syntactic unit X°” (Law 2006, p. 647). This condition is evidentially based on suppletive forms such as de el = del ‘of the’ and a el = al ‘to the’ in Spanish, as well as a les = aux ‘to the’ in French, con lo = collo ‘with the’ in Italian, an dem = am ‘at the’ in German, and so on. Note that it is the incorporation that allows for the suppletion, not the other way around. As Law argues, this syntax-morphology-interface condition does not always require suppletion for the syntactic unit to be formed; if a language has such forms at all, it is evidence that all determiners incorporate, with or without suppletion. As such, although Spanish suppletion is not robust, consisting solely of del ‘of the’ and al ‘at the’, Law argues that all Spanish D+P constructions (e.g., de la ‘of the’, a la ‘at the’, con el ‘with the’, etc.) involve incorporation and must syntactically operate as one unit. For wh-movement, for example, the only option then is to move the entire Spanish PP, as the wh-element has formed a syntactic unit with the preposition, as shown in (11).

(11)	Araceli	no	sabe	[PP [P+D con quéi]]	[DP ti hombre]]j	Rosario
	Araceli	not	know.3s	with-what	man	Rosario
	está	comiendo tj.
	is	eating
	‘Araceli doesn’t know with what guy Rosario is eating.’

English, lacking any such suppletive forms, does not have the same syntax-morphology interface condition of D+P incorporation. Therefore, in English wh-movement, for example, the wh-element can be freely extracted since it can be syntactically separated from the preposition, as shown in (12a); additionally, this lack of incorporation does not prevent optional pied-piping, as shown in (12b).

(12)

a.

Emma doesn’t know [DP what guy]i Jen is eating [PP [P with] ti].

b.

Emma doesn’t know [PP [P with [DP what guy]]]i Jen is eating ti.

2.2. Heritage Speaker Bilingualism

A heritage language can be defined as “a language spoken at home or otherwise readily available to young children … [that] is not a dominant language of the larger (national) society” (Rothman 2009, p. 156). Although heritage grammars and monolingual grammars are similar in that they are acquired via naturalistic input at a young age, substantial differences have been found between them (e.g., Montrul 2008; Silva-Corvalán 1994; among others). Numerous linguistic phenomena have been found to manifest differently in the grammars of heritage speakers, with specific examples from Spanish including subject and object expression (Montrul 2004), subject-verb inversion (Cuza 2013), and presentational focus (Hoot 2017), to name a few.

It is important to note that heritage speakers, “unlike mature monolingual speakers who ultimately converge into a common steady state, exist in a linguistic continuum with no clear or easily measurable cut-off points” (Pascual y Cabo and Gómez Soler 2015, p. 188). Although any group of speakers is heterogeneous by nature, acquiring two languages instead of one exponentially increases the potential for differences among the language backgrounds, meaning that a meaningful degree of variation among such bilinguals (even from the same speech community) is not merely a possibility but an inevitability.

Age of acquisition (AoA) is just one of many variables that has been used to better understand within-group differences for heritage language speakers. Essentially, researchers can look at whether the point at which an individual is exposed to the dominant societal language results in different linguistic outcomes for a given grammatical property (Arnaus et al. 2021; Schulz and Grimm 2019; Tsimpli 2014). Although the specific age range varies throughout the literature, the term simultaneous bilingual is commonly used for an individual who is exposed to both languages from birth (or a very early age) while the term sequential bilingual refers to someone “who first acquired one language, then a second language after the rudiments of the first language were established” (Montrul 2008, p. 94). This classification is just a starting point, as the term sequential bilingual can encompass a wide variety of individuals. As such, some researchers have made even further distinctions, for example, delineating early sequential bilinguals (AoA 5;0) and child sequential bilinguals (AoA 3;0) (Meisel 2013; Montrul 2008). For the current study and its focus on US heritage speakers of Spanish, the AoA of English is used to determine if an individual is either a simultaneous or child sequential bilingual, as that is the language where age of exposure will vary more considerably.

2.3. P-Stranding in Heritage Speaker Spanish

One specific syntactic phenomenon that has been shown to diverge between monolinguals and heritage speakers of Spanish is p-stranding. Interestingly, Depiante and Thompson (2013) found evidence that US heritage speakers consistently rated p-stranding as more acceptable in Spanish than a Spanish-dominant control group of recent immigrants. Given this data, it would suggest that there would be no potential conflict for p-stranding in Spanish–English code-switching for these individuals. However, Pascual y Cabo and Gómez Soler (2015) found evidence that “the English structure bleeds into [the] Spanish” (p. 203) of some, but importantly not all, heritage speakers. In their study, they targeted con ‘with’ and en ‘in’ in simple wh-questions, embedded wh-questions, and relative clauses across three experimental tasks: (a) a judgment task with p-stranding in Spanish sentences (n = 30), (b) a judgment task with pied-piping in Spanish sentences (n = 30), and (c) a production task with ‘dehydrated’ Spanish sentences4 (n = 10). They included two different experimental groups that differed according to US heritage speakers’ AoA of English. The first group comprised bilingual speakers (n = 21) who learned both Spanish and English from birth, whereas the second group comprised child sequential bilinguals (n = 12) who did not learn English until after age 6. Both groups had intermediate/advanced proficiency in Spanish, as well as comparable self-rated proficiencies. A comparison group of Spanish native speakers (n = 11) born and raised in Mexico until at least the age of 16 also participated in the study.

The findings show, first, that p-stranding in Spanish was rejected by the comparison monolingual group, whereas it was possible for (some) heritage speakers. Crucially, the two heritage speaker groups diverged, with child sequential bilinguals exhibiting the expected rejection of p-stranding in Spanish, which mirrored the comparison group of monolingual speakers. The simultaneous bilinguals, however, showed no restriction for p-stranding in Spanish, suggesting that the construction has been extended from English for these individuals, as it is “a domain of grammar vulnerable to crosslinguistic influence during the formative years” (Pascual y Cabo and Gómez Soler 2015, p. 203). In short, when it comes to p-stranding in heritage speaker Spanish, there is evidence of a categorical difference based on AoA of English.

2.4. Research Questions and Hypotheses

The current study adopts a constraint-free approach to CS, and it is agnostic to whether it needs to be a traditional minimalist approach to CS that is lexicalist (MacSwan 1999, 2014, 2021) or exoskeletal (Grimstad et al. 2018). Both are within the generative minimalist program (Chomsky 1995), but differ regarding whether syntactic structures are generated from features on lexical items (i.e., lexicalist) or generated by features independently of lexical items (i.e., exoskeletal). Under such an approach, constraints are due to the interaction of the two grammars in question, specifically when there is a mismatching of features. This mismatching mirrors exactly what happens in monolingual derivations; in other words, this approach assumes there is no ‘third grammar’ that is unique to CS.

Within this framework, the following research question and sub-question are formulated:

(13)	Do heritage speakers of Spanish accept p-stranding in Spanish–English CS, and if so, does the AoA of English play a role?

Based on previous research, we can formulate specific hypotheses to these questions: yes, heritage speakers of Spanish should find p-stranding acceptable in CS; and yes, the results should vary based on the age of onset of bilingualism. Assuming the simultaneous bilinguals replicate the findings of Pascual y Cabo and Gómez Soler (2015) and accept p-stranding in Spanish, and assuming the construction is syntactically parallel across the two grammars (i.e., they have English-like non-incorporation of D+P across the board), there should be no restriction whatsoever against p-stranding in CS for these speakers. If they allow p-stranding monolingually in both of their languages, following Law’s (2006) analysis, the DP can be freely extracted as an independent syntactic unit (regardless of whether an English DP is extracted from a Spanish PP or vice versa). However, the same cannot be said for the child sequential bilinguals. Assuming they reject p-stranding in Spanish, this would be evidence of D+P incorporation, thus exposing an asymmetry between their Spanish and English grammars. Consequently, there should also be a restriction on p-stranding in CS, with three different possibilities. The child sequential bilingual CS results will depend on which element(s) motivate(s) D+P incorporation: the determiner, the preposition, or both. Specifically, the three options are: (a) if incorporation is dependent only upon the features inherent to the determiner, p-stranding should be only accepted with English-to-Spanish switches (i.e., an English DP with a Spanish preposition); (b) if incorporation is dependent only upon the features inherent to the preposition, p-stranding should only be accepted with Spanish-to-English switches (i.e., a Spanish DP with an English preposition); or (c) if incorporation is dependent upon the features of both the determiner and the preposition, then p-stranding should be rejected in all switch cases.

To summarize, simultaneous bilinguals are hypothesized to accept p-stranding in monolingual English and Spanish, as well as both CS scenarios (i.e., Spanish-to-English and English-to-Spanish). Child sequential bilinguals, however, are hypothesized to reject p-stranding in monolingual Spanish, accept it in monolingual English, and at least some (if not all) p-stranding in CS will be rejected.

3. Methods

3.1. Procedure

The central component of the experiment was a written5 acceptability judgment task (AJT) that used a 7-point Likert scale. Each sentence was presented one at a time with the question How acceptable is this sentence? or ¿Qué le parece esta oración? ‘How is this sentence to you?’. The points on the scale were labeled as follows: 1 = completely unacceptable/completamente inaceptable, 2 = mostly unacceptable/mayormente inaceptable, 3 = somewhat unacceptable/un poco inaceptable, 4 = unsure/no sé, 5 = somewhat acceptable/un poco aceptable, 6 = mostly acceptable/mayormente aceptable, 7 = completely acceptable/completamente aceptable. While the monolingual blocks used their respective questions and labels, the CS block used the Spanish question and English labels in an effort to enhance bilingual mode.

Completed entirely online via the software Qualtrics (Qualtrics, LLC, Provo, UT, USA), participants saw a sentence, clicked their response on the scale, and then clicked an arrow button in the lower right portion of the screen to advance to the next stimulus. The AJT was untimed, meaning participants could take as much time as they liked to complete these steps for each item, and the stimuli were completely randomized within each block for every participant. An example of what the AJT looked like for the participants is shown in Figure 1.

The AJT was embedded within a larger experimental procedure that included five sections. First, participants completed a consent form and a brief task training. The training included a general overview of what AJTs ask an individual to do (i.e., use their intuitions, avoid prescriptive rules, etc.) while also including several practice sentences. Following González-Vilbazo et al. (2013), this first section also primed participants for bilingual mode via its use of Spanish–English CS throughout the instructions (with all switches occurring at the clausal boundary or with adjunct phrases). Next, participants completed an entire block of CS judgments. The third section of the procedure switched to monolingual Spanish mode, with a multiple-choice cloze test proficiency measure (Montrul and Slabakova 2003) and then an entire block of Spanish judgments6. Afterward, the participants switched to monolingual English mode, which included a different multiple-choice cloze test proficiency measure (O’Neill et al. 1981) followed by a block of monolingual English judgments. Finally, participants wrapped up the procedure with a background questionnaire.

3.2. Participants

The individuals who participated in the study were all US heritage speakers of Spanish (N = 29), categorized as such by virtue of being second-generation immigrants whose home language was a minority language (i.e., Spanish), as well as being more fluent in the dominant societal language (i.e., English) (Polinsky and Kagan 2007; Valdés 2000). Some participants were removed from the dataset (n = 5) for either not being a self-reported code-switcher and/or for indicating a negative attitude towards CS (Badiola et al. 2018). The remaining participants (n = 24) were between 19 and 49 years old (M = 23.2) and were either born in the US (n = 20) or arrived as a young child (M = 4.8 years). Importantly, all participants learned both Spanish and English from a young age.

Following Pascual y Cabo and Gómez Soler (2015), the participants were divided into two different groups based on their AoA of English. The simultaneous heritage speakers reported learning both languages at about 1 year of age, with the cut-off being before age 5 (n = 13), whereas the sequential heritage speakers learned English at age 5 or later (n = 11). This division in the participants was chosen to align with Pascual y Cabo and Gómez Soler (2015), thereby making the results directly comparable. An overview and comparison of the two groups’ general linguistic profiles is provided in

Table 1. Participant overview.

Factor	Simultaneous Bilinguals		Child Sequential Bilinguals
	M	SD	M	SD
Age of acquisition:
English	1.3	1.6	5.5	1.5
Spanish	0.3	0.5	0.3	0.5
Proficiency score 1:
English (out of 40)	37.1	1.9	36.4	2.0
Spanish (out of 50)	36.9	6.2	37.6	4.7
Self-rated proficiency 2:
English (out of 5)	5.0	0.0	4.9	0.3
Spanish (out of 5)	4.5	0.5	4.5	0.7
English dominance 3 (out of ±218)	31.3	43.0	20.0	25.1

¹ Proficiency was measured via two separate multiple-choice cloze tests for English (O’Neill et al. 1981) and Spanish (Montrul and Slabakova 2003). ² The scale for self-rated proficiency was as follows: 1—Poor; 2—Needs work; 3—Good; 4—Very good; 5—Native speaker command. ³ Language dominance was measured via the Bilingual Language Profile (Birdsong et al. 2012).

.

3.3. Stimuli

All target sentences included in the AJT contained p-stranding (N = 32). Half of the stimuli included embedded wh-movement and half included relative clauses. There were 8 tokens for each of the 4 language conditions: monolingual English, monolingual Spanish, English-to-Spanish CS, and Spanish-to-English CS. Examples of the embedded wh-movement p-stranding stimuli for each language condition are presented in (14), and examples of the relative clause p-stranding stimuli are in (15).

(14)

a.

Bill doesn’t know what woman Megan is arguing with.

	b.	Manuel	no	sabe	qué	señora	Ximena	está	discutiendo
		Manuel	not	know.3s	what	lady	Ximena	is	arguing
		con.
		with
		‘Manuel doesn’t know what lady Megan is arguing with.’

	c.	Bill doesn’t know what woman	Ximena	está	discutiendo	con.
			Ximena	is	arguing	with
		‘Bill doesn’t know what woman Ximena is arguing with.’

	d.	Manuel	no	sabe	qué	señora	Megan is arguing with.
		Manuel	not	know.3s	what	lady
		‘Manuel doesn’t know what lady Megan is arguing with.’

(15)

a.

Lucy is the girl that Gabe is going out with.

	b.	Leticia	es	la	chica	que	Arturo	está	saliendo	con.
		Leticia	is	the	girl	that	Arturo	is	going-out	with
		‘Leticia is the girl that Arturo is going out with.’

	c.	Lucy is the girl	que	Arturo	está	saliendo	con.
			that	Arturo	is	going out	with
		‘Lucy is the girl that Arthur is going out with.’

	d.	Leticia	es	la	chica	that Gabe is going out with.
		Leticia	is	the	girl
		‘Leticia is the girl that Gabe is going out with.’

Filler stimuli with various other types of constructions (and switches for the CS conditions) were also included (N = 169). These sentences primarily targeted adverb order, auxiliary verbs, and pronouns.7 Examples of monolingual filler stimuli are shown in (16–17), and examples of CS filler stimuli are shown in (18–19). The full set of stimuli included in the experiment can be found in Appendix A.

(16)

a.

*

Sarah buys frequently tomatoes.

b.

Sarah frequently buys tomatoes.

(17)	a.	Margarita	compra	frecuentemente	tomates.
		Margarita	buys	frequently	tomatoes
		‘Margarita frequently buys tomatoes.’

	b.	Margarita	frecuentemente	compra	tomates.
		Margarita	frequently	buys	tomatoes
		‘Margarita frequently buys tomatoes.’

(18)	a.	*	Sus	hermanos	han	eaten pizza every day.
			Their	brothers	have
			‘Their brothers have eaten pizza every day.’

(17)	b.	Sus	hermanos	have eaten pizza every day.
		Their	brothers
		‘Their brothers have eaten pizza every day.’

(19)	a.	*	Ayer	ellos	bought some peaches.
			Yesterday	they
			‘Yesterday they bought some peaches.’

	b.	Ayer	esos	hombres	bought some peaches.
		yesterday	those	men
		‘Yesterday those guys bought some peaches.’

4. Results

The acceptability scores provided by participants were first standardized, converting them into z-scores. The mean ratings are presented in Figure 28, separated out by the two participant groups and four different language conditions. Recall that the higher end of the scale was labeled “acceptable,” and as such a more positive z-score is indicative of a sentence’s increased acceptability. Note, though, that there is no finite point on the y-axis that we can label as “acceptable” or “unacceptable”; what we are interested in rather is comparing which structures were found to be more acceptable or less acceptable relative to each other.

If we look first at just the ratings for the monolingual stimuli, we see the expected pattern. The results for the child sequential bilinguals clearly show the crosslinguistic asymmetry, with much higher acceptability ratings for p-stranding in English (M = 0.39, SD = 0.60) when compared to Spanish (M = −1.20, SD = 0.82). The results for the simultaneous bilinguals are similar in that they also show a preference for p-stranding in English (M = 0.40, SD = 0.44), with nearly identical ratings to those of the child sequential bilinguals; however, the two groups diverge in that the simultaneous bilinguals did not seem to outright reject p-stranding in Spanish in that they rated it much higher than the child sequential bilinguals, with scores hovering much closer to the middle of the scale (M = −0.33, SD = 0.93).

Turning to the CS results, the availability of p-stranding in mixed sentences is essentially a mirror image of the monolingual results. The crosslinguistic asymmetry of p-stranding for child sequential bilinguals carries over to the code-switched stimuli, as they provided much lower acceptability scores for p-stranding in English-to-Spanish CS (M = −0.78, SD = 0.87), which as a reminder is exemplified in (14c) and (15c), as compared to Spanish-to-English (M = 0.27, SD = 0.94), as exemplified in (14d) and (15d). The simultaneous bilinguals similarly preferred p-stranding when the switch occurred from Spanish-to-English (M = 0.27, SD = 0.92); however, like in the monolingual results, they did not outright reject p-stranding in English-to-Spanish switches like the other group as their ratings were much higher (M = −0.27, SD = 1.11).

A two-way ANOVA was conducted to investigate the effect of bilingual type and language(s) on z-score, and a significant interaction was found, F(3,375) = 7.777, p < 0.001. Post hoc analysis revealed that both groups showed a significant asymmetry for the monolingual stimuli, rating p-stranding in English more favorably than in Spanish (p < 0.001). Importantly, though, the groups also differed from each other in that the simultaneous bilinguals were overall more accepting of p-stranding than child sequential bilinguals, regardless of the language (p < 0.001). Post hoc analysis also confirmed the parallel nature of the monolingual and CS results; that is, there was no significant difference found between the English and Spanish-to-English ratings (p > 0.005), nor was there a significant difference between the Spanish and the English-to-Spanish ratings (p > 0.05).

Before concluding the results section, it is worthwhile to explore AoA a bit more in detail. Although the expected difference was found between the two groups of heritage speakers of Spanish, recall that the simultaneous bilinguals did not accept p-stranding to the same degree in both of their languages, as a significant asymmetry was still found between their Spanish and English p-stranding. This is a bit surprising if we understand these bilinguals to have transferred English-like p-stranding to their Spanish. If their grammars have aligned for this construction, why are they not more uniformly aligned? To investigate this further, we can modify the way AoA of English is operationalized. Instead of dividing the participants into two separate groups with a specific age as a cut-off point, we can plot each participant’s mean z-score for each language condition by their self-reported AoA of English9. This configuration of the results is presented in Figure 3. Here we can see that the same overall pattern holds. Note that p-stranding in English and in English-to-Spanish CS received stable, higher acceptability scores, regardless of the specific AoA of English of the participants. Meanwhile, p-stranding in both Spanish and Spanish-to-English CS showed a steady decline in acceptability as the AoA of English increased. When analyzed this way, we see that there was not a categorical divide regarding AoA, but rather a continuum. This point should not be surprising for researchers in bilingualism, but it is important to reiterate that although terms like simultaneous and sequential are helpful in discussing important differences in bilingual grammars, we need to constantly be mindful of heterogeneity within these groups.

Overall, the results show that the availability of p-stranding in Spanish–English CS follows directly from whether a Spanish or English preposition can be stranded monolingually by a bilingual speaker. That is to say, the language of the preposition indicates whether p-stranding is available or not. Specifically, English prepositions allow p-stranding for any monolingual or mixed sentence regardless of the type of heritage speaker, while Spanish prepositions allow for it only if the speaker is a simultaneous bilingual.

5. Discussion

When it comes to the availability of p-stranding in Spanish–English intrasentential CS, the current study shows that it is possible, but both AoA of English and switch direction can mitigate its acceptability. As argued by Pascual y Cabo and Gómez Soler (2015), for heritage speakers of Spanish, p-stranding “is a domain of grammar that can be affected developmentally depending on the (socio)linguistic realities in which these individuals are immersed during the first years of acquisition” (p. 203). Syntactically speaking, the results from this experiment provide evidence that the Spanish of simultaneous bilinguals is influenced by English in that there is no D+P incorporation. This creates a parallelism between their Spanish grammar and their English grammar, which not only allows for free extraction of DPs from PPs in monolingual contexts but also permits p-stranding when mixing their two grammars, with no restriction on CS (regardless of the direction of the switch). In short, if a simultaneous Spanish–English bilingual wishes to strand a preposition, they can do so when using English, Spanish, or any mixture of the two.

This finding for the simultaneous bilinguals raises an interesting issue. If it is indeed the case that there is no D+P incorporation for such individuals, this would suggest that there would never be any pied-piping with prepositions for them. Since the current study only tested p-stranding, we cannot make concrete claims as to whether the simultaneous bilinguals do or do not optionally employ pied-piping in English, Spanish, and/or code-switching instead of p-stranding. Moreover, Law’s (2006) proposal does not account for the optionality of pied-piping in English. Unlike in languages like Spanish where pied-piping is derived from the combination of wh-movement and the D+P incorporation, it is unclear what motivates optional pied-piping in English. Nevertheless, whatever accounts for it in monolingual speakers can be extended to the simultaneous bilinguals. That is to say, it may be the case that simultaneous bilinguals have drawn on some sort of optionality for p-stranding, as in English, based on the interplay of properties of D and P for English and Spanish.

Contrasting with the simultaneous bilinguals, the child sequential bilinguals show an asymmetry that is in line with what the literature has shown regarding monolingual speakers. Essentially, when it comes to p-stranding, their Spanish grammar is given time to develop the crucial syntactic restriction (i.e., D+P incorporation) before there is enough influence from English.10 Interestingly, this asymmetry has specific consequences when it comes to mixing the two grammars in the same sentence. For child sequential bilinguals, extraction of a Spanish DP out of an English PP is acceptable, but not vice versa. This asymmetry is evidence that the D+P incorporation in their Spanish grammar presents itself in CS, but only in specific contexts. Recall that it was hypothesized that there would be a restriction on p-stranding in CS for child sequential bilinguals, but that it would be one of three different possibilities depending on the specific element(s) that initiates the D+P incorporation. Since p-stranding is available in Spanish-to-English switches and only in such switches, these results suggest that it is the language of the preposition and not the determiner that determines incorporation.

A lingering question is the extensiveness of the lack of D+P incorporation in Spanish for the simultaneous bilinguals. Recall that the availability of p-stranding in both CS contexts points to the absence of a unified syntactic unit, which is essential to Law’s (2006) argument for how p-stranding is prohibited in languages like Spanish. However, also recall that the only preposition tested here was with/con. This form does not have any suppletive forms in Spanish, so all we can say for certain is that there is evidence that D+P incorporation does not occur with con ‘with’ for simultaneous Spanish–English bilinguals. It seems implausible to argue that there is no D+P incorporation at all for these speakers; although it was not tested here, we can presume that these speakers do employ the two Spanish suppletive forms, del ‘of the’ and al ‘at the’. Assuming they do, future research should test whether p-stranding with the prepositions de ‘of’ and a ‘to’ behave the same way as con ‘with’ for heritage speakers of Spanish, or if there is perhaps variability depending on the particular preposition. It is possible that there is only a syntax-morphology-interface condition for de ‘of’ and a ‘to’, whereas all other Spanish prepositions (which lack any suppletive forms) are free to be stranded for these bilinguals.

Although the primary focus of this study is CS, it is important to explicitly acknowledge that the results provide additional syntactic evidence for the behavior of p-stranding in both Spanish and English. First, the current results replicate those of Pascual y Cabo and Gómez Soler (2015) for heritage speakers of Spanish, as a parallel difference was found between simultaneous and child sequential bilinguals. Additionally, the current results also provide more evidence of the general, ubiquitous availability of p-stranding in English.

6. Limitations and Conclusions

There are various limitations to the current study. As described earlier, the experiment only included one specific preposition, with/con, and as such we should not assume that the results here are representative of the way all prepositions behave regarding p-stranding in Spanish–English CS. There is known idiosyncratic variation with p-stranding depending on the preposition (Biber et al. 1999), so it should be expected that this would carry over to CS as well. Additionally, the data were obtained from an AJT, so it is unclear if production data would show the same patterns. It is possible that some participants are hesitant to accept p-stranding when providing a judgment, but then actually use p-stranding in their everyday speech without an issue (or vice versa). Finally, as already mentioned, heritage language speakers (like bilinguals more generally) are a heterogeneous group regarding their linguistic backgrounds, and here we only looked closely at their AoA of English. Continued investigation of other linguistic variables would likely provide more detailed information about the behavior of p-stranding in Spanish–English CS.

Nonetheless, the current study provides a clear first step toward understanding when p-stranding is available in CS. Since simultaneous Spanish–English bilinguals’ grammars allow p-stranding in both languages, there is no restriction in CS. These results suggest there is no D+P incorporation in their grammars, allowing for free extraction of DPs from PPs. However, the asymmetry found for child sequential bilinguals shows that extraction of a Spanish DP out of an English PP is acceptable, but not vice versa. These findings are evidence that they have D+P incorporation in Spanish, which presents itself in switched contexts sometimes. Specifically, these results suggest that it is the preposition and not the determiner that determines incorporation, and as such the language of the preposition dictates whether there is p-stranding or not.