Productive Vocabulary Knowledge Predicts Acquisition of Spanish DOM in Brazilian Portuguese-Speaking Learners

Abstract

Differential object marking (DOM) in Spanish refers to the overt morphological marking of certain direct objects. Specifically, this a-marking of direct objects is driven by animacy and usually precedes human objects. Other features such as specificity and definiteness matter to a lesser extent. This direct object a-marking has also been attested in Brazilian Portuguese (BP), but with more restrictions than in Spanish. Thus, BP is typically not considered a DOM language. This article discusses the acquisition of DOM in second language (L2) Spanish among BP speakers. Seventy-four adult Brazilians with various levels of L2 proficiency completed three experimental tasks: elicited production, acceptability, and productive vocabulary knowledge, which measures productive lexical knowledge and was designed specifically to assess this language pair. We analyzed our data to uncover the knowledge that BP-speaking L2 learners of Spanish possess of DOM and to determine whether their knowledge of DOM can be predicted by their L2 productive vocabulary knowledge. Results indicate that the learners acquire the animacy-driven nature of Spanish DOM, both in their productive and receptive grammatical knowledge, as their productive vocabulary knowledge increases. The L2 learners present different acquisitional outcomes yet those with deeper productive vocabulary knowledge show a stronger animacy-driven distribution of DOM in Spanish.

1. Introduction

In this article, we discuss the acquisition of differential object marking (DOM) in second language (L2) Spanish among Brazilian Portuguese (BP) adult speakers. We examine the use and acceptability of DOM, and we pay special attention to the role of grammaticality, animacy, and productive vocabulary knowledge (used as a proxy for measuring overall language proficiency). In Spanish, DOM is determined by properties of the direct object, mainly animacy and definiteness, marking these arguments morphologically with the so-called personal a. The same type of case-marking also exists in BP, but DOM in this language is subject to many more restrictions than in Spanish (Cyrino and Irimia 2019) and thus, it is overall significantly less frequent than in Spanish.

The study’s goal is (i) to uncover the type of knowledge that BP native speakers acquiring L2 Spanish possess of DOM in their L2, and (ii) to determine whether their knowledge of DOM in the L2 is modulated by their L2 productive vocabulary knowledge. Our research questions and hypotheses are guided by previous proposals on second language acquisition: the Feature Reassembly Hypothesis (Lardiere 2008, 2009), the Missing Surface Inflection Hypothesis (Haznedar and Schwartz 1997; Prévost and White 2000), and the Full Transfer/Full Access Hypothesis (Schwartz and Sprouse 1996). These previous works allow us to hypothesize on the acquisition process that L1 BP speakers go through when acquiring DOM in Spanish, namely the role of L1 transfer (e.g., omission and overextension) and variability in morphological competence in production. Second, we use a picture naming task, PLEP (Prueba Léxica de Español y Portugués/Prova Léxica de Espanhol e Português), which is a novel task designed by the research team to measure productive lexical knowledge of Spanish-BP bilinguals, which we use as a proxy for overall L2 proficiency, as suggested by previous research (e.g., Bedore et al. 2012; Gollan et al. 2012; Sheng et al. 2014; Treffers-Daller and Korybski 2015). Thus, we want to test whether productive vocabulary knowledge can predict the acquisition of the animacy-driven distribution of DOM in Spanish among L1 BP speakers. For all the tasks in the study, we also collected data in the participants’ L1.

The organization of the article is as follows: Section 2 introduces the literature on DOM in Spanish and in BP, followed by a summary of studies on the acquisition of DOM in L2 Spanish, and then, an overview of lexical knowledge as a proxy for overall proficiency. Section 3 presents the research questions and hypotheses. Next, Section 4 provides the methodology, including information about the participants and data collection, as well as the data analysis. Section 5 presents the results from the tasks in Spanish and BP. Finally, Section 6 and Section 7 contain the discussion and conclusion, respectively.

2. Literature Review

2.1. Differential Object Marking in Spanish and BP

Differential object marking (DOM) is a well-studied phenomenon in Spanish, in which transitive verbs mark their arguments morphologically with the a. The use of Spanish DOM is determined by the properties of the direct object, especially when the direct object carries the lexical feature of animacy and the discourse pragmatic feature of definiteness (Aissen 2003; Bossong 1991; Leonetti 2004, 2008; Torrego 1998). The examples below illustrate the Spanish DOM in the presence/absence of animacy and definiteness. That is, animacy is the prominent feature (Leonetti 2008), as it usually precedes human objects (1a) (e.g., la hija), as opposed to inanimate objects (1b) (e.g., la casa). In cases where the direct object is an animal (e.g., la vaca), García García (2018) argues that DOM is optional (1c).

(1a)	El	padre	cuida	a	la hija.	[+animate, +definite]
	The	father	takes care	of	DOM the daughter
	‘The father takes care of the daughter.’

(1b)	El	padre	cuida	Ø	la	casa.	[−animate, +definite]
	The	father	takes care	of	the	house
	‘The father takes care of the house.’

(1c)	El	padre	cuida	(a)	una vaca.	[+animate, −definite]
	The	father	takes care	of	(DOM) a cow
	The father takes care of a cow.’

Some exceptions of the [+animate, +definite] rule for DOM are indefinite objects that denote humans since they can receive DOM as in 1(d) (Leonetti 2004), indefinite pronouns that denote human beings such as nadie ‘nobody’ are mandatorily marked with a despite of this pronoun being [-definite] 1(e), which indicates that animacy is the main trigger for DOM use in Spanish. Additionally, a is also used as a disambiguating mechanism when both the subject and the object are inanimate (1f) (Rodríguez-Mondoñedo 2006).

(1d)	El	padre	busca (a)	una niñera	[+animate, −definite]
	The	father	looks for	(DOM) a nanny
	‘The doctor looks for a “nanny”’

(1e)	El	detective	no encontró	a nadie	[+animate, −definite]
	The	detective	did not find	DOM anyone
	‘The detective did not find anyone.’

(1f)	El	ácido	corroe	al metal.	[−animate, +definite]
	The	acid	corrodes	DOM the metal
	‘The acid corrodes the metal.’

According to Wiskandt (2021), DOM mechanisms that depend on the properties of the object itself are language-specific and are arranged on a scale based on animacy, definiteness/specificity, or on a combination of both. Determining when DOM is overtly expressed, depends on whether the objects are ranked high on such scales. Take, for example, Aissen (2003, p. 437), who proposed two scales that arrange animacy and definiteness features in a hierarchical form and explained that DOM is overtly expressed when the objects are ranked high in such scales.

(2a)

Animacy scale: human > animate > inanimate

(2b)

Definiteness scale: personal pronoun > proper name > definite NP > indefinite specific NP > non-specific NP

A-marking of direct objects has also been attested in Brazilian Portuguese (BP). Some similarities across these two languages have been noted mainly because the features of animacy, definiteness and specificity seem to respect the same hierarchical order of these constraints (Schwenter 2014). Nevertheless, DOM in BP is subject to many more restrictions than in Spanish (Cyrino and Irimia 2019), to the extent that BP is typically not considered a DOM language.

For instance, in BP, DOM occurs in cases of clitic doubling, where DOM precedes tonic personal pronouns in direct object position (3a), but nominal direct objects in simple transitive sentences do not receive DOM in BP, regardless of the object being animate and specific (Wiskandt 2021), as in (3b).

(3a)	Ana me	abraça	a mim.
	Ana	hugs	DOM me
	‘Ana hugs me.’

(3b)	Ana	abraça	o menino.
	Ana	hugs	Ø the boy.
	‘Ana hugs the boy.’

Cyrino (2017) also proposed that when two human direct objects appear next to each other due to verb ellipsis, the second one received a-marking for grammaticality and interpretation purposes (4a). In addition, Cyrino and Irimia (2019) proposed that in sentences where it is possible to omit the overt a-marker, it is often used to disambiguate meaning. (4a) and (4b) illustrate the change in meaning generated by the presence of DOM, which generates the reading that the subject Ana’s action of hugging affected two objects: the boy and the grandfather. In contrast, (4c) may license a reading in which both Ana and the grandfather hugged the boy. The lack of DOM before the word avô ‘grandfather’ can make it dubious if the grandfather is a second object or is functioning as the subject of a coordinated clause.

(4a)	Ana	abraça	o menino	e	ao avô
	Ana	hugs	the boy	and	DOM the grandfather
	‘Ana hugs the boy and the grandfather’.

(4b)	Ana	abraçou	o menino	e	ao avô também
	Ana	hugged	the boy	and	DOM the grandfather too
	Ana hugged the boy and the grandfather too.

(4c)	Ana	abraçou	o menino	e	o avô também
	Ana	hugged	the boy	and	Ø the grandfather too.

In summary, objects are overtly marked in Spanish depending on their lexico-semantic features, usually when they are animate, while they are usually unmarked if they are inanimate. BP, on the other hand, is not typically considered a DOM language, as objects in BP are only overtly marked in restricted contexts: when preceding tonic personal pronouns in direct object position and when used to disambiguate the syntactic function of a noun group in sentence presenting an elided verb. Overall, both languages share the morphology of this accusative case marking yet they differ in their distribution: it is ruled mostly by the lexico-semantic features of the object in Spanish while it depends on syntactic contexts in BP.

2.2. The Acquisition of DOM in L2 Spanish

DOM is acquired very early on in Spanish L1 acquisition (Rodríguez-Mondoñedo 2008). As for L2 Spanish, some learners face difficulties even at the intermediate and advanced levels of proficiency, while others seem to successfully develop a robust knowledge of this syntactic phenomenon.

Previous studies have shown that, for L2 learners whose L1 is English, a language that does not have DOM, this structure presents serious challenges, and even though instruction facilitates development, native-like performance is hard to attain. For example, Guijarro-Fuentes and Marinis (2007) used an acceptability judgment task to examine the sensitivity to the semantic features relevant to the use of DOM in Spanish such as animacy, specificity/definiteness, and telicity, among others. The authors concluded that L2 learners have difficulties acquiring structures involving the syntax/semantics interface and that it is only at the advanced level when sensitivity to the least complex DOM constraints (animacy and specificity/definiteness) begins to develop. Such study examined learners at the low-intermediate, high-intermediate, and advanced levels. In turn, Guijarro-Fuentes (2012) observed the acquisition of interpretable features in L2 learners of Spanish by investigating the accusative a-marking in Spanish. This study examined the production and comprehension of DOM using a Completion Task and an Acceptability Judgment Task across different proficiency groups. Similarly to Guijarro-Fuentes and Marinis (2007), all groups of L2 learners (intermediate-low, intermediate-high, and advanced levels) presented different grammatical knowledge of DOM in Spanish from that of native speakers. The author concluded that even though some interpretable features, such as animacy, are acquirable relatively early, others present developmental effects. Other studies (e.g., Bowles and Montrul 2008, 2009; Farley and Wiebe 2004; Wiebe 2003) have analyzed the effects of different types of instructional intervention and/or feedback, and have found that while there is improvement, a native-like knowledge of DOM is not attained.

Other researchers have investigated the Spanish DOM among speakers whose L1 has DOM (e.g., Romanian and Turkish), but with some degree of variation in comparison to Spanish. For example, Montrul and Gürel (2015) examined the production, comprehension, and acceptability of DOM in Spanish among low-intermediate and intermediate Turkish-speaking L2 learners. Turkish is a DOM language that is definiteness-driven. The findings revealed that both groups produced DOM with animate direct objects, but the intermediate learners outperformed the low-intermediate ones. Comprehension and acceptability data showed knowledge and sensitivity to grammaticality regarding DOM in Spanish. In another study, Montrul (2019) examined the acquisition of DOM in Spanish among advanced and near-native speakers of Romanian, a language presenting a DOM that is primarily triggered by referential stability (definiteness), whereas Spanish DOM is animacy-driven. Using production, comprehension, and acceptability tasks, the study revealed that the learners performed similarly to Spanish native speakers across the three tasks, suggesting that they have successfully aligned the features [+animate] and [+specific] to DOM in Spanish, despite their L1 having a different hierarchical order for these features. A more recent study by López Otero (2022) investigated the production and acceptability of DOM in Spanish by Romanian sequential bilinguals. An elicited production task showed that Romanians use DOM in Spanish with animate objects very frequently, but they also produce DOM, to a lesser extent, with inanimate pronoun objects. Acceptability data showed that these bilinguals accepted animate objects with no DOM, as well as inanimate objects with DOM.

To our knowledge, Yokota (2001) is the first study that examined the acquisition of DOM in the L2 Spanish of BP speakers. The author collected L2 Spanish production data via several methods: an open-ended written narrative task, a written translation task, and a written elicited production task. Over one hundred participants featuring a wide range of instruction hours in Spanish (from below 60 to more than 250) completed these tasks in different sessions. Yokota identified several acquisition stages of DOM in the L2 learners: omission, overextension, substitution for prepositions as well as possible reinterpretation of DOM as a part of the prepositional regime of some verbs. The author determined that low-proficient speakers produced some instances of DOM in their Spanish and advanced speakers produced target-like DOM around 75–80% of the time, depending on the task. Additionally, Dos Santos and Cunha Silva (2005) brought forward the difficulty that this phenomenon poses to BP-speaking L2 learners of Spanish. In a study aiming to discern what L2 Spanish errors were more easily identified by other learners, the authors found that DOM was omitted by several learners yet spotted as an error by other learners. However, L2 speakers mostly detected errors involving verb morphology and spelling mistakes related to accent use.

In sum, the literature has shown that the use of DOM in an L2 is a challenging task. However, speaking an L1 presenting a DOM can play a facilitative role in acquiring DOM in L2 Spanish even if the L1 DOM is ruled by lexical-semantic features different from animacy, such as definiteness. BP L1 speakers, the group of learners examined in the present study, also experience difficulties acquiring DOM in their L2 Spanish in spite of BP presenting accusative a-marking in some restricted syntactic contexts and of the similarities between BP and Spanish in terms of lexicon and functional morphology. The goal of the present study is to examine the acquisition of DOM in L2 Spanish by speakers of BP, a language that presents a restricted DOM, while exploring the role of a proficiency continuum operationalized by a productive vocabulary task, which we seek to validate as a reliable tool to assess overall proficiency in the Spanish-BP language pair, following Gollan et al.’s (2012) MINT.

2.3. Lexical Knowledge as a Proxy for Overall Proficiency

Language ability is multidimensional, which makes using a single measure to assess different language components challenging, whether it is vocabulary or any other domain (see He 2019; Richards and Malvern 2007). As noted by Gollan et al. (2012), there is still no standard method for determining language proficiency; however, previous research (Bates and Goodman 1997) has suggested that, in any given language, the diversity of linguistic forms relies almost entirely on the lexicon, and even L1 grammatical development depends, to a large extent, on vocabulary knowledge. In L1 acquisition, vocabulary size is a major predictor of language proficiency (Bates and Goodman 1997), and so it is in the second language context.

Several researchers have emphasized the potential relationship between vocabulary knowledge and performance in the four macro skills (i.e., speaking, writing, listening, and reading) (e.g., Alharthi 2018, 2020; Crossley et al. 2012; Laufer and Aviad-Levitzky 2017; Qian and Lin 2019; Wang and Treffers-Daller 2017; Zhang and Zhang 2020). More specifically, as noted by Milton (2013), this predictiveness is enhanced with a combination of measures of different dimensions of lexical knowledge. For instance, Zhang and Zhang (2020) found a positive correlation between L2 productive and receptive vocabulary knowledge and L2 reading/listening comprehension, while Alharthi (2020) found that productive vocabulary knowledge correlates with and can predict speaking success. Additionally, Crossley et al. (2012) took a computational approach to the study of lexical knowledge and provided supporting evidence for the strength of lexical indices to determine L2 competence. The authors analyzed lexical knowledge including breadth (which refers to word frequency and lexical diversity), depth, and accessibility in written texts, and found that the aforementioned indices correctly classified 70% of the texts based on proficiency level. The authors also suggest that this measure would be effective across languages with important implications for the assessment and classification of lexical knowledge and L2 proficiency.

Even though vocabulary tests can be an easily quantifiable direct measure of proficiency (Milton 2009), vocabulary knowledge is a highly complex construct (Treffers-Daller 2019). According to He (2019), in the last 30 years, researchers have undertaken the task of creating a variety of vocabulary size measures or tests. For instance, even within the realm of vocabulary knowledge, there is still a distinction between receptive and productive knowledge. To name a few examples, the Peabody Picture Vocabulary Test (Dunn and Dunn 1981), the Vocabulary Levels Test (Nation 1983, 1990; Schmitt et al. 2001), the Mill Hill Vocabulary Scale (Raven 1943), and the Lextale-Esp (Izura et al. 2014) focus on receptive vocabulary size, while the Productive Levels Test (Laufer and Nation 1999) and the Multilingual Naming Test (Gollan et al. 2012) assess productive knowledge. Other considerations include whether a test focuses on breadth versus depth (Read 1993).

Many vocabulary knowledge tests address lexical frequency, usually dividing this variable into three categories (He 2019): high-frequency words, mid-frequency words, and low-frequency words. To determine the frequency level of a word, researchers have different resources available such as The General Service List (West 1953) which contains a list of roughly 2000 words that represent the most frequent words of written English, and for other languages, the Corpus del Español (Davies 2018a), and Corpus do Português (Davies 2018b), which provide billions of words of recent data from different Hispanic and Lusophone countries. For example, recent studies have used lexical range, sophistication, and richness in oral production tasks to determine language dominance in bilingual populations (Daller et al. 2003; Gollan et al. 2012; Treffers-Daller and Korybski 2015). In addition, when designing/adopting a test, according to Wesche and Paribakht (1996), its purpose, the criteria for selecting the target vocabulary items, how the test is presented to the test-taker, and the format of the responses need to be clearly understood. Another fundamental consideration is age appropriateness, and as noted by Gollan et al. (2012) whether the test was designed for monolingual or bilingual populations.

We included this literature on lexical knowledge measures because, in the present study, a productive lexical test was used to assess L2 proficiency. More specifically, we see this type of measure suitable for the phenomenon under investigation, DOM, based on the previous evidence suggesting that grammatical development is linked to vocabulary knowledge. For this study, we developed a productive vocabulary test specifically designed to measure BP-Spanish bilinguals. Our test was inspired by the aforementioned Multilingual Naming Test (Gollan et al. 2012); however, it contains different test items that are presented in a specific order based on lexical frequency. We provide further details about our test in the Methods section. The next two sections of this paper, Research Questions, and Methods, will further address the rationale and the composition of the lexical test we implemented.

3. Research Questions

The research questions of the present study are guided by three proposals on L2 acquisition: Lardiere’s (2008, 2009) Feature Reassembly Hypothesis (FRH), the Full Transfer/Full Access hypothesis (FT/FA; Schwartz and Sprouse 1996), and the Missing Surface Inflection Hypothesis (MSIH; Haznedar and Schwartz 1997; Prévost and White 2000). The FRH assumes that acquiring a language is a process in which learners select a subset of features among all the features available across languages and match those features with morphology. Consequently, L2 learners face the task of identifying and selecting the features in their L2, which may be available in their L1, and matching those features to the morphology in their L2. Lardiere (2008, 2009) claims that morphological competence is crucial in this process as learners can have identified and selected the appropriate features but may find difficulties in matching them to the target-like morphology or in producing such morphology.

The MSIH (Haznedar and Schwartz 1997; Prévost and White 2000) also emphasizes the difficulty that L2 learners encounter when acquiring functional morphology, particularly in production. This hypothesis suggests that, even though L2 learners have underlying knowledge of functional categories and features, they often lack the surface morphology in L2 and that is why they are absent in production. Both the FRH and the MSIH claim that differences between production and comprehension in L2 learners can stem from performance issues in production.

The FT/FA hypothesis (Schwartz and Sprouse 1996) can help predict patterns in L2 acquisition. Specifically, the FT/FA hypothesis proposes that the abstract properties of the L1 grammar (e.g., features and functional categories) get transferred to the L2 first (Full Transfer), and then, when learners encounter input data that cannot be handled by their L1 alone, they are forced to restructure their grammar drawing from options of UG (Full Access). By helping establish a claim regarding initial stages of L2 acquisition, the FT/FA hypothesis complements the FRH.

In the present study, we examine the acquisition of DOM in BP-speaking L2 learners of Spanish. This group of learners needs to acquire that direct objects are overtly marked in some contexts, particularly if they are animate. These learners’ L1 is not considered a DOM language as direct objects are overtly marked in very restricted syntactic contexts (Cyrino and Irimia 2019). However, both languages use the same morphology for these different accusative case markers (i.e., a). In other words, BP-speaking L2 learners of Spanish must reassemble [animacy] and [definiteness] into the accusative marker a. This reassembly process, specifically for [animacy] given the animacy-driven nature of DOM in Spanish, motivates our research questions:

Research Question 1: What knowledge do Brazilian Portuguese-speaking L2 learners of Spanish possess of DOM in their second language?

Hypothesis 1.

We hypothesize that BP-speaking L2 learners of Spanish reassemble animacy features into their L2 DOM, as per productive and receptive grammatical knowledge. This reassembly process occurs after an initial stage of L1 transfer followed by variability in their productive knowledge. Following the FRH, their acquisition process consists of reassembling animacy features into their L2 DOM. Specifically, animate objects are overtly marked whereas inanimate objects are not. This process can lead to several stages presenting variable expressions of DOM, particularly in the L2 learners’ production as opposed to their comprehension, as a result of the difficulty in developing morphological competence in the L2, as identified by both the FRH and the MSIH. Additionally, following the FT/FA hypothesis, we hypothesize that the L2 learners go through an initial stage of L1 transfer featuring omission of DOM or use of non-animacy-driven a-marking in their L2 Spanish prior to starting the feature reassembly process. In sum, we predict our results to show a rather consistent acquisition pattern in the L2 learners’ comprehension data while presenting variability in their production. Previous studies, namely Yokota (2001), found several stages in the production of DOM in the L2 Spanish of BP speakers, some of which featured omission and overextension. With regard to stages of acquisition, we predict L1 transfer at initial stages, which could encompass omission of DOM as well as non-animacy-driven use of DOM in their L2 Spanish, leading to overextension.

Research Question 2: Is knowledge of DOM in L2 Spanish modulated by productive L2 vocabulary knowledge among BP speakers?

Hypothesis 2.

We hypothesize that acquisition of DOM in their L2 is tied to their productive vocabulary knowledge in Spanish. In other words, productive vocabulary knowledge may indicate whether the participants in this group of L2 learners have successfully reassembled animacy features into DOM in Spanish as well as account for within-group variability. This hypothesis is consistent with previous studies that have documented a correlation between overall proficiency and productive vocabulary knowledge in bilinguals (e.g., Bedore et al. 2012; Gollan et al. 2012; Sheng et al. 2014; Treffers-Daller and Korybski 2015). This second hypothesis builds on our first hypothesis in two ways: first, given that both the FRH and the MSIH claim that morphological competence in production is challenging for L2 learners, we expect productive vocabulary knowledge to account for the L2 learners’ variable productive knowledge of DOM in their L2 Spanish. Second, following the FT/FA hypothesis, we predicted L1 transfer at initial stages, which could encompass omission of DOM as well as non-animacy-driven use of DOM in their L2 Spanish, leading to overextension. Therefore, with regard to acquisition stages, we expect productive vocabulary knowledge to indicate that the L2 learners with more limited productive vocabulary knowledge are still going through this initial L1 transfer stage, characterized by omission or overextension of DOM. On the other hand, those L2 learners with deeper productive vocabulary knowledge will demonstrate more animacy-driven knowledge of DOM in Spanish than their counterparts whose productive vocabulary knowledge in Spanish is more limited.

4. Methods

4.1. Participants

74 BP-speaking L2 Spanish adult speakers (54 female and 20 male; mean age = 33.84, SD = 9.87) participated in this study. They were all born and raised in Brazil, where they were still living at the moment of data collection. All participants were native speakers of BP who had learned Spanish in Brazil as a second language (age of onset of acquisition range = 7–45; M = 19.23; SD = 7.47; age of onset of L2 fluency range = 14–48; M = 25.34; SD = 7.60; 9 participants reported not having reached fluency in Spanish). The participants spoke varieties of BP from all the regions in Brazil, being the Southeast and Northeast regions the most represented, followed by the South, Central-West, and North.

Table 1. States and regions represented in the participant pool.

Region	States
Southeast (n = 37)	Rio de Janeiro, São Paulo, Minas Gerais, Espírito Santo
Northeast (n = 17)	Maranhão, Piauí, Ceará, Rio Grande do Norte, Paraíba, Pernambuco, Alagoas, and Bahia
South (n = 10)	Paraná, Santa Catarina, Rio Grande do Sul
Central-West (n = 2)	Distrito Federal
North (n = 1)	Tocantins

provides the list of states that were represented in our participant pool based on regions.

The study used a recruitment flyer that contained the researchers’ contact information, a brief general description of the study, the requirements, and the participant compensation. The Consejería de Educación en Brasil, which is the institution that represents the Spanish Ministry of Education and Professional Training in Brazil, agreed to help disseminate the recruitment flyer on their website, and people who were interested contacted the researchers.

The study employed a purposive sampling technique in order to achieve comparability among the participant pool, but we did not target specific proficiency levels. Thus, we anticipated variability among the participants given the uniqueness of their bilingual experience (e.g., diverse learning backgrounds and levels of exposure to Spanish). A comparison group of monolingual or Spanish-dominant speakers was not necessary because the focus of this study is not to determine how similar or different BP-speaking L2 learners of Spanish are from native speakers. Our focus was on the acquisition of the animacy-driven distribution of DOM in the learners’ L2 Spanish, particularly in SVO sentences, within a proficiency continuum, as well as on whether their productive vocabulary knowledge could predict their knowledge of DOM in Spanish. Additionally, the performance of comparison groups in previous studies examining the animacy-driven distribution of Spanish DOM in SVO sentences usually presents ceiling effects and a lack of variability (e.g., Guijarro-Fuentes and Marinis 2007; Hur 2022; López Otero 2020, 2022, inter alia).

4.2. Data Collection Instruments and Procedures

Prior to beginning the study, participants read and signed a consent form, which informed them of the length of the study, the procedures, and confidentiality of their data, as well as the compensation for their time (a R$50 gift card for a store in Brazil, which is the equivalent of a 10 USD gift card). All the gift cards were electronic codes that were distributed to the participants via email upon completion of the study.

The first part of the study was a meeting on Zoom with one of the researchers to sign the consent form and complete the first task which involved a lexical knowledge test. Lexical knowledge can be used as a proxy for general language proficiency (Gyllstad 2016), so we designed and administered the Prueba Léxica de Español y Portugués/Prova Léxica de Espanhol e Português (PLEP), which is a novel picture naming task that specifically assesses Spanish-BP bilinguals (see Appendix A). Using a traditional grammar test alone with BP-Spanish bilinguals is problematic because the syntactic and morphological proximity between this language pair could lead to ceiling effects. Furthermore, adopting another well-established lexical test such as the MiNT would not be ideal due to the high number of cognates between the languages. The PLEP examines the knowledge of nouns (similar to the MiNT) and can be administered in both languages. It contains 58 items that are not cognates in the language pair. The test items are organized based on lexical frequency; that is, the lexical frequency counts of the items in each language were matched (as much as possible) using the frequency data provided in the Corpus del Español (Davies 2018a), and the Corpus do Português (Davies 2018b).

For this study, we administered the PLEP first in Spanish, then in BP, and the scores are reported in

Table 2. Scores of the PLEP in Spanish and BP and Spanish DELE.

	Spanish PLEP	BP PLEP	Spanish DELE
Mean	34.96	56.45	45.2
Standard deviation	11.80	2.13	3.76

, which shows that, as expected, the participants’ lexical knowledge in BP is stronger and less variable than in Spanish, the L2. In addition, we administered a version of the DELE (Diploma de Español como Lengua Extranjera) test with 50 items (Cuza et al. 2013; Montrul and Slabakova 2003). The DELE assessed morphosyntax and lexicon, both of which these two languages may share, and as shown in Table 2, there was a ceiling effect. 71 out of 74 participants scored over 40/50. Thus, we do not consider the DELE to be an optimal instrument for this language pair.

The participants completed the PLEP on Zoom and their participation was recorded for further analysis. Correct responses were coded as ‘1’ while incorrect responses or a lack thereof were coded as ‘0’. Like in the MINT (Gollan et al. 2012), participants were offered conceptual and phonemic cues, yet this was not considered in our analysis. All the participants completed the PLEP in Spanish, their L2, first, followed by BP, their first and dominant language. The PLEP was administered in such order to prevent priming effects from their dominant into their non-dominant language. The ceiling effects in their BP PLEP scores confirm that the results in their dominant language are not affected by completing the PLEP in this specific order while their Spanish PLEP scores are a reflection of the large range of proficiency levels captured in the participant pool. We acknowledge, nonetheless, that an alternative order might have impacted the results, as indicated by a reviewer. After completing the PLEP on Zoom, the participants were given their participant code and a Qualtrics link containing the DELE test, a language background questionnaire (LEAP-Q; Marian et al. 2007), and four experimental tasks: two written Elicited production Tasks (EPT) and two Acceptability Judgment Tasks (AJT). The Spanish EPT and AJT were completed, followed by the BP EPT and AJT (see Appendix B).

The EPT was used to determine whether the participants’ use of DOM was animacy-driven or not. The role of definiteness, however, was not explored as all direct objects were definite. In this task, they had to type the correct response to complete the sentence. The direct objects were given in parenthesis, and in the Spanish EPT, all direct objects were feminine, while in the BP EPT they were all masculine. All the verbs and direct objects featured in the Spanish and BP versions of the EPT were cognates. The EPTs had 16 experimental items in each language, distributed in two conditions (k = 8): [+animate], for which a-marking is expected in Spanish, and [-animate], for which no accusative marking is expected in Spanish.

Table 3. Spanish and BP EPT item samples per condition.

Animacy	Spanish Test Item Samples
Animate	Mariana está alegre porque su hijo agradeció ________ (la enfermera). ‘Mariana is joyful because her son thanked ________ (the nurse).’
Inanimate	Juana está feliz porque…su hijo vio ________ (la revista). ‘Juana is happy because her son saw ________ (the magazine).’
Animacy	BP Test Item Samples
Animate	Francisco está tranquilo porque seu filho obedeceu _______ (o médico). ‘Francisco is tranquil because his son obeyed ________ (the doctor).’
Inanimate	Joana está feliz porque seu filho viu ________ (o filme). ‘Joana is happy because her son saw ________ (the film).’

below shows some EPT item samples.

The AJTs examined the participants’ receptive grammatical knowledge of DOM under a variable affecting the direct object in Spanish: [±animate]. The AJTs comprised 16 experimental items (48 distractors per language) distributed in four conditions (k = 4; 16 in each language). The experimental items were the same as in the EPT yet presented in an AJT format: specifically, the participants were presented with sentences (and their corresponding preambles and prompts) containing either animate or inanimate direct objects in grammatical and ungrammatical constructions (controlled for by the use or non-use of DOM), and were asked to rate each of them using a Likert scale from 1 to 5 (1 = Completamente extraño ‘Completely odd’, 2 = Extraño ‘Odd’, 3 = Ni bien ni mal ‘Neither good nor bad’, 4 = Bien ‘Well’, 5 = Completamente bien ‘Completely well’), which was present below every test item. Every time a participant rated an item 1 or 2, they were also asked to explain why, in order to confirm that they had rejected the item due to the phenomenon under examination, which in this case was object marking. The items were counterbalanced and pseudorandomized. The preamble and the prompts were equivalent in the two languages; the main difference between the tasks in each language is that the Spanish AJT presented feminine direct objects and the BP AJT featured masculine direct objects. This decision was taken in order to facilitate the perception of the a-marking and to avoid bound morphology (i.e., al (a + el) for Spanish and à (a + a) in BP).

Table 4. Spanish and BP AJT item samples per condition.

Animacy	Gram.	Spanish Test Item Samples
Animate	Gram.	Manuela está orgullosa porque… Su hijo escuchó a la profesora.
	Ungram.	Ana está triste porque… * Su hijo empujó Ø la señora.
Inanimate	Gram.	José está preocupado porque… Su hijo escuchó Ø la canción vulgar.
	Ungram.	Beatriz está triste porque… * Su hijo empujó a la mesa.
Animacy	Gram.	BP Test Item Samples
Animate	Gram.	Felipe está feliz porque… Seu filho viu Ø o rei.
	Ungram.	João está furioso porque… * Seu filho insultou ao professor.
Inanimate	Gram.	Cristina está surpresa porque… Seu filho admirou Ø o quadro.
	Ungram.	Beatriz está triste porque… * Seu filho empurrou ao móvel.

In linguistics, * means that the sentence is not grammatical.

shows the distribution of items across conditions. All tasks were preceded by instructions and four practice items.

4.3. Data Analysis

The data from the EPT were analyzed using two models, one for the Spanish and one for the BP production data. In both models, DOM (use = 1, non-use = 0) was the dependent variable, while animacy of the direct object (animate or inanimate) was the independent variable. The model analyzing the Spanish EPT model included productive vocabulary knowledge as measured by the Spanish PLEP as an additional independent variable, which was continuous. Additionally, the Spanish EPT model explored the interaction between animacy and Spanish PLEP scores. Both models included random intercepts for each subject as well as for each lexical item.

The data from the AJT, on the other hand, were analyzed by using two ordinal regression models, one in each language. In both models, response (1 through 5) was the dependent variable while grammaticality (grammatical or ungrammatical), and animacy of the object (animate or inanimate) were the independent variables. As in the EPT, the Spanish AJT model also included Spanish PLEP scores as an independent variable. Furthermore, the Spanish AJT model tested the interaction between grammaticality and PLEP scores while the BP AJT model included the interaction between condition and animacy. In all models, significance of main effects and all possible interactions were assessed using hierarchical partitioning of the variance via nested model comparisons.

Before analyzing the data, we discarded responses that were unrelated to the acquisition of DOM. In the Spanish EPT, from 1184 total responses, 173 were discarded. Most of these discarded responses, some of which we present below in bold, included uses of prepositional phrases (e.g., su hijo agradeció por la oportunidad ‘their son thanked for the opportunity’) instead of use or non-use of DOM as well as direct objects featuring different lexical items than those targeted in the task (e.g., su hijo obedeció a la autoridad ‘their son obeyed DOM the authority’) as well as other structures that resulted in ungrammatical sentences when combined with the prompt (e.g., su hijo insultó la insultó ‘their son insulted her’; su hijo admiró una mujer muy guapa passó y su hijo admiró ‘their son admired one very beautiful woman passed by and their son admired’). Four out of 74 participants had all their Spanish EPT responses discarded as their responses fell into the categories described above (presenting the following Spanish PLEP scores: A14 = 32/58; J21 = 39/58; J55 = 43/58; J552 = 20/58). In the BP EPT, from 1184, 104 responses were discarded. These responses also featured prepositional phrases (e.g., Seu filho agradeceu pelo momento especial ‘their son thanked for the special moment’), different structures not including the direct object provided in the task (e.g., seu filho o escutou ‘their son listened to him’). Some responses from the AJT data were also discarded: 14 out of the 1184 responses of the Spanish AJT and 27 out of the 1184 responses of the BP AJT. These responses were rejections that participants justified with answers unrelated to the use of DOM, mostly related to the use of prepositional phrases in both their Spanish (e.g., tocó en la princesa ‘touched the princess’) and BP (e.g., tocou no príncipe ‘touched the prince’).

5. Results

5.1. Results from Spanish Data

5.1.1. Elicited Production Task

The EPT results revealed that the BP-speaking L2 learners of Spanish use DOM, particularly to mark animate objects. More specifically, they marked animate objects 51.4% of the time (SD = 0.50); however, they also overextended their L2 DOM to inanimate objects in some instances (M = 0.24; SD = 0.43). The generalized linear mixed model examining the Spanish EPT data found that animacy was a predictor for the L2 learners’ responses (β = −1.18, SE = 0.47, z = −2.51, p = 0.01). However, their productive vocabulary knowledge as measured by the PLEP was not (β = 0.01, SE = 0.01, z = 1.12, p = 0.26). Additionally, the model found an interaction between animacy and PLEP scores (β = −0.01, SE = 0.01, z = −2.06, p = 0.04), which indicates that the participants produced more instances of DOM in animate contexts as their PLEP scores increased. This interaction is illustrated in Figure 1 below.

Individual Analysis Attending to Developmental Stages

In this section, we analyzed the participants’ patterns of responses with the goal of confirming the group effects described above and of discerning any potential developmental stages in their productive knowledge. We classified the participants’ EPT responses into seven categories according to Yokota’s (2001) findings: (1) omission: no a-marking was used in any test item, (2) overextension: a-marking was used in all test items, (3) some non-animacy-driven a-marking: less than half of the test items received a-marking and such marking was distributed across animate and inanimate objects, (4) incipient animacy-driven marking: animacy-driven use of a-marking yet presenting more than 4 mistakes out of 16 test items (in this category, if inanimate objects were marked, animate objects were marked at least twice as much), (5) almost target-like use of DOM: animacy-driven use of DOM presenting no more than 4 mistakes out of 16 test items, (6) target-like use of DOM: animate objects received a-marking categorically and inanimate objects did not receive any marking, and (7) other responses: as described in Section 4.2, four participants had all their EPT responses discarded. These seven categories, the number of participants within each of them and their PLEP scores are presented in

Table 5. Types of marking in the Spanish EPT in increasing average PLEP score order.

Marking	Description	No. of Participants	Category-Based Group Object Marking of Animate Objects	Category-Based Group Object Marking of Inanimate Objects	Category-Based Group PLEP Scores
Overextension	A-marking in most test items	n = 17	M = 6.86/8; SD = 1.35	M = 5.43/8; SD = 1.39	Range = 8–50; M = 33.29; SD = 12.72
Other responses	Discarded responses	n = 4	N/A	N/A	Range = 20–43; M = 33.5; SD = 10.08
Omission	No instances of DOM use	n = 15	M = 0.00/8; SD = 0.00	M = 0.00/8; SD = 0.00	Range = 19–51; M = 33.6; SD = 9.26
Incipient animacy-driven marking	Animacy-driven DOM with more than 4 mistakes. If participants marked inanimate objects, they marked animate objects at least twice as much.	n = 15	M = 3.70/8; SD = 1.30	M = 0.71/8; SD = 0.87	Range = 11–49; M = 33.6; SD = 12.04
Some non-animacy-driven marking	Objects are marked less than half the time and regardless of animacy	n = 10	M = 2.03/8; SD = 1.48	M = 1.70/8; SD = 1.24	Range = 14–52; M = 34.67; SD = 12.97
Almost target-like use of DOM	Animacy-driven DOM with up to 4 mistakes	n = 11	M = 7.20/8; SD = 0.83	M = 1.8/8; SD = 0.90	Range = 20–55; M = 40.55; SD = 11.96
Target-like use of DOM	Animacy-driven DOM	n = 2	M = 8.00/8; SD = 0.00	M = 0.00/8; SD = 0.00	Range = 42–44; M = 43; SD = 1.41

below.

We decided to establish a distinction between the categories ‘incipient animacy-driven marking’ and ‘almost target-like use of DOM’ by drawing a comparison between our data and Yokota’s (2001) data, as she found that advanced BP-speaking L2 learners of Spanish produced target-like DOM in their L2 around 75–80% of the time, which, in our dataset, would represent 12 out of 16 test items. Therefore, the ‘almost target-like use of DOM’ category included participants who made up to 4 mistakes in their EPT. This cut-off line is consistent with previous experimental research on the acquisition of Spanish DOM in other bilingual populations (e.g., Montrul and Sánchez-Walker 2013).

The results above suggest that the majority of L2 learners do not have knowledge of the animacy-driven distribution of DOM in Spanish. A look at the average Spanish PLEP scores of each potential developmental stage indicates that most groups present similar productive vocabulary knowledge of their L2 language, except for those who possess target-like or almost target-like productive knowledge of this structure. While these results confirm the suitability of productive vocabulary knowledge as a reliable predictor for grammatical competence in L2 learners, they are insufficient to determine the order in which the developmental stages shown above occur.

Individual Analysis Attending to Verbs Tested

In this section, we analyzed the participants’ EPT responses with the goal of exploring any potential item effects, particularly to rule out the possibility that participants may have interpreted the a-marking as a prepositional regime of any given verb.

Table 6. DOM use across test items and conditions in the Spanish EPT.

Verb	Animate Objects	Inanimate Objects
Admirar ‘admire’	M = 0.45/1; SD = 0.50	M = 0.19/1; SD = 0.40
Agradecer ‘thank’	M = 0.63/1; SD = 0.49	M = 0.23/1; SD = 0.43
Empujar ‘push’	M = 0.45/1; SD = 0.50	M = 0.22/1; SD = 0.42
Escuchar ‘listen’	M = 0.56/1; SD = 0.50	M = 0.30/1; SD = 0.46
Insultar ‘insult’	M = 0.53/1; SD = 0.50	M = 0.19/1; SD = 0.40
Obedecer ‘obey’	M = 0.60/1; SD = 0.49	M = 0.32/1; SD = 0.47
Tocar ‘touch’	M = 0.43/1; SD = 0.50	M = 0.25/1; SD = 0.44
Ver ‘see’	M = 0.45/1; SD = 0.50	M = 0.19/1; SD = 0.40

below provides mean and standard deviation values for each verb in the two conditions under examination (i.e., animate vs. inanimate direct objects) for all the participants in the study.

The data shown in the table above indicate that, overall, participants produced considerably more instances of DOM with animate objects than with inanimate objects. Indeed, these data seem to indicate that DOM is used twice as often with animate objects than with inanimate ones. These results suggest that we can reject the possibility of learners erroneously interpreting DOM in their L2 as a prepositional verb regime. Additionally, in order to rule out any potential reinterpretation of DOM in this group of L2 learners, a further analysis was carried out with a focus on the group of participants who produced some a-marking yet not following an animacy-driven distribution. The results of said analysis are shown in

Table 7. DOM use across test items and conditions within the participants who produced some non-animacy-driven a-marking in the Spanish EPT.

Verb	Animate Objects	Inanimate Objects
Admirar ‘admire’	M = 0.30/1; SD = 0.48	M = 0.10/1; SD = 0.32
Agradecer ‘thank’	M = 0.50/1; SD = 0.53	M = 0.11/1; SD = 0.33
Empujar ‘push’	M = 0.20/1; SD = 0.42	M = 0.13/1; SD = 0.35
Escuchar ‘listen’	M = 0.11/1; SD = 0.33	M = 0.33/1; SD = 0.50
Insultar ‘insult’	M = 0.30/1; SD = 0.48	M = 0.29/1; SD = 0.49
Obedecer ‘obey’	M = 0.33/1; SD = 0.50	M = 0.30/1; SD = 0.48
Tocar ‘touch’	M = 0.13/1; SD = 0.35	M = 0.20/1; SD = 0.42
Ver ‘see’	M = 0.11/1; SD = 0.33	M = 0.22/1; SD = 0.44

below.

The similar values found throughout some verbs in Table 7 above suggest that some participants might have reinterpreted DOM in their L2 as a prepositional verb regime, specifically with insultar ‘insult’ (M = 0.30 for animate vs. M = 0.29 for inanimate) and obedecer ‘obey’ (M = 0.33 for animate vs. M = 0.30 for inanimate). However, we identified the only three participants within this participant subgroup that all of whom produced some instances of non-animacy-driven a-marking, who produced a-marking in both the animate and inanimate object conditions with the same verb: A17 (Spanish PLEP score: 30/58) produced a-marking with both escuchar ‘listen’ and obedecer ‘obey’, J35 (Spanish PLEP score: 41/58) made use of a-marking with obedecer ‘obey’, and J42 (Spanish PLEP score: 39/58) produced a-marking with empujar ‘push’ as well. In conclusion, while finding only three participants who marked both animate and inanimate objects with the same verb might suggest that these three participants may have reinterpreted a-marking with those specific verbs, it rejects the possibility that such reinterpretation extends throughout the L2 learners.

In summary, the individual analysis of the production data indicates that this group of L2 learners features different acquisitional outcomes, which may represent stages of an acquisition continuum. The L2 speakers showing an animacy-driven production of DOM are those who, on average, have the highest values of productive vocabulary knowledge. An item analysis revealed that this group of L2 learners does not reinterpret a-marking as a prepositional verb regime, except for three learners.

5.1.2. Acceptability Judgment Task

Additionally, the AJT data suggest that the L2 learners can discern between grammatical and ungrammatical instances of use and non-use of DOM in Spanish. Figure 2 shows the L2 learners’ responses to the Spanish AJT across grammaticality and animacy, while

Table 8. Distribution of acceptance of grammatical and ungrammatical items with mean and standard deviation according to animacy in the Spanish AJT.

Grammaticality	Animacy	Response
Grammatical	Animate	M = 4.26/5; SD = 1.04
Grammatical	Inanimate	M = 4.15/5; SD = 1.06
Ungrammatical	Animate	M = 3.50/5; SD = 1.33
Ungrammatical	Inanimate	M = 3.56/5; SD = 1.34

below provides averages and standard deviations of their responses.

The ordinal regression examining the Spanish AJT data determined that grammaticality is a predictor for the L2 learners’ responses (β = −1.01, SE = 0.11, t = −9.02, p < 0.01) while neither animacy nor PLEP scores were when examining their potential effects. However, this ordinal regression also found an interaction between grammaticality and PLEP scores (β = −0.45, SE = 0.11, t = −4.14, p < 0.01), which indicates that the L2 learners’ responses to ungrammatical test items were modulated by their productive vocabulary knowledge, as shown in Figure 3 below.

5.2. Results from BP Data

The BP EPT data indicate that the L2 learners of Spanish mark objects in their L1. Particularly, they marked more animate objects (M = 0.21; SD = 0.41) than inanimate ones (M = 0.123; SD = 0.33). However, a generalized linear mixed model did not find this contrast to be significant (β = −0.87, SE = 0.88, z = −0.98, p = 0.33).

Table 9. A-marking use across test items and conditions in the BP EPT.

Verb	Animate Objects	Inanimate Objects
Admirar ‘admire’	M = 0.10/1; SD = 0.30	M = 0.04/1; SD = 0.21
Agradecer ‘thank’	M = 0.56/1; SD = 0.50	M = 0.13/1; SD = 0.34
Empurrar ‘push’	M = 0.06/1; SD = 0.24	M = 0.01/1; SD = 0.12
Escutar ‘listen’	M = 0.18/1; SD = 0.38	M = 0.19/1; SD = 0.40
Insultar ‘insult’	M = 0.13/1; SD = 0.34	M = 0.10/1; SD = 0.30
Obedecer ‘obey’	M = 0.56/1; SD = 0.50	M = 0.36/1; SD = 0.48
Tocar ‘touch’	M = 0.10/1; SD = 0.30	M = 0.03/1; SD = 0.18
Ver ‘see’	M = 0.03/1; SD = 0.17	M = 0.11/1; SD = 0.32

below provides mean and standard deviation values for each verb in the two conditions under examination (i.e., animate vs. inanimate direct objects) for all the participants in the study.

Regarding the BP AJT data, the participants also gave different responses to grammatical and ungrammatical test items. This suggests that BP-speaking L2 learners of Spanish reject the object marker in their L1, as expected. Figure 4 shows the L2 learners’ responses to the BP AJT across grammaticality and animacy.

Table 10. Distribution of acceptance with mean and standard deviation based on animacy and grammaticality in the BP AJT.

Grammaticality	Animacy	Response
Grammatical	Animate	M = 4.40/5; SD = 0.86
Grammatical	Inanimate	M = 4.49/5; SD = 0.79
Ungrammatical	Animate	M = 3.77/5; SD = 1.36
Ungrammatical	Inanimate	M = 3.44/5; SD = 1.45

below provides averages and standard deviations of their responses.

An ordinal regression examining the BP AJT data confirmed that their responses were modulated by grammaticality (β = −0.79, SE = 0.16, t = −5.03, p < 0.01). Additionally, this regression found an interaction between grammaticality and animacy (β = −0.68, SE = 0.22, t = −3.02, p < 0.01), which suggests that the L2 learners rate the ungrammatical presence of the object marker with inanimate objects lower than with animate objects, as seen in Figure 4 and Table 9 above.

5.3. Summary of the Results

To sum up, the Spanish language results indicate that the L2 learners under examination acquire the animacy-driven nature of Spanish DOM as their productive vocabulary knowledge increases. This applies to both their productive knowledge and their receptive grammatical knowledge. The L2 learners present different acquisitional outcomes yet those with deeper productive vocabulary knowledge have knowledge of the animacy-driven distribution of DOM in Spanish. Additionally, an item analysis revealed that, with the exception of three learners, DOM was not reinterpreted as a prepositional verb regime. Overall, these findings support the validity of the tool used to measure their productive vocabulary knowledge, the PLEP. This finding indicates that the PLEP is a more suitable proficiency measurement tool than the DELE for this population: the PLEP can predict DOM knowledge in the L2 Spanish of BP-speaking L2 learners while ceiling effects were found in the participants’ responses to the DELE.

The BP results, on the other hand, indicate that the participants reject the use of an object marker in their language. In addition, they reject marked inanimate objects at a higher rate than marked animate objects. Additionally, they produced some instances of object marking in the EPT.

6. Discussion

The present study investigated the acquisition of DOM in the L2 Spanish of L1 BP speakers. Following previous studies examining the acquisition of DOM in L2 Spanish from a feature reassembly lens (e.g., Curione 2021; López Otero 2020, 2022; Mayer and Sánchez 2021; Montrul and Gürel 2015; Nediger et al. 2016, inter alia), our first research question and hypothesis were guided by Lardiere’s (2008, 2009) FRH. Additionally, our first research question and hypothesis were also informed by the MSIH (Haznedar and Schwartz 1997; Prévost and White 2000) and the FT/FA hypothesis (Schwartz and Sprouse 1996). We hypothesized that BP-speaking L2 learners of Spanish are capable of reassembling the animacy features into their L2 DOM. Specifically, this reassembly process is preceded by an L1 transfer stage, which presents both omission and overextension of DOM, and by variable productive knowledge of DOM, as opposed to receptive grammatical knowledge. This hypothesis was confirmed by the L2 learners’ data. The L2 learners’ production data indicates that BP-speaking L2 learners of Spanish go through an initial stage of L1 transfer in which they either omit or overextend DOM in their L2 Spanish, consistently with the FT/FA hypothesis. This initial stage is followed by the beginning of the feature reassembly process, which presents variable expressions of DOM, as predicted by both the FRH and the MSIH. Finally, some L2 learners feature productive knowledge of animacy-driven DOM in their L2. Additionally, their receptive grammatical knowledge of Spanish DOM is sensitive to grammaticality. Therefore, these results illustrate the reassembly process involved in the acquisition of DOM in L2 Spanish among L1 BP speakers and its successful acquisition in some of the L2 learners. It is consistent with the hypotheses discussed above with regard to acquisition stages and to the contrast between variable productive knowledge versus more stable receptive grammatical knowledge as a result of the difficulties in performance with functional morphology.

Our second research question inquired as to whether the L2 learners’ knowledge of DOM was modulated by their L2 productive vocabulary knowledge. Following previous studies reporting on correlations between overall proficiency and productive vocabulary knowledge in bilinguals (e.g., Bedore et al. 2012; Gollan et al. 2012; Sheng et al. 2014; Treffers-Daller and Korybski 2015, inter alia), we hypothesized that the L2 learners with deeper productive vocabulary knowledge in their L2 reassemble animacy features into DOM morphology of their L2 successfully, as opposed to those learners presenting a rather limited productive vocabulary knowledge, who are still in the process of reassembling the aforementioned features. This hypothesis built on our first hypothesis and claimed that productive vocabulary knowledge can account for the progress throughout the acquisition stages posited by our first hypothesis. Our results confirm our second hypothesis. The results from the inferential statistics indicate that the L2 learners’ EPT and AJT responses are predicted by their productive knowledge. Specifically, the L2 learners with higher Spanish PLEP scores produced more instances of animacy-driven DOM. From the AJT, in turn, we learned that higher vocabulary knowledge led to a more robust knowledge of Spanish DOM, as those L2 learners with deeper productive vocabulary knowledge gave lower ratings to ungrammatical test items. Additionally, the operationalization of the acquisition stages formulated in our first hypothesis and how these stages relate to the L2 learners’ productive vocabulary knowledge also confirm our second hypothesis: the group of participants who featured total omission or total overextension of DOM in their L2 Spanish presented lower PLEP scores than those who presented variable productive knowledge of DOM. In turn, the group of L2 learners who showed an animacy-driven use of DOM presented the highest average PLEP score.

These findings suggest that the syntactically restricted accusative marker in BP, which is constrained to very limited syntactic contexts, does not play a significant facilitative role in the acquisition of DOM in L2 Spanish. Indeed, although direct objects introduced by the verbs agradecer and obeceder received a-marking more than half of the time in the BP EPT, the initial stages for the acquisition of DOM in L2 Spanish among L1 BP speakers appear to show L1 transfer effects, which take form as either omission or overextension of DOM. This finding is consistent with the view that BP is not a DOM language (Cyrino and Irimia 2019). Nevertheless, most L2 speakers in this study seem to be aware of the existence of a-marking in Spanish: only 15 participants did not produce any instances of DOM in their EPT. However, these results are not comparable with those found in studies examining the acquisition of DOM by L2 learners whose first languages present DOM systems as well: intermediate Turkish-speaking and Romanian-speaking L2 learners of Spanish showed a more animacy-driven distribution of Spanish DOM than most of the participants in the present study (Montrul 2019; Montrul and Gürel 2015), as target-like and almost target-like productive knowledge of DOM in Spanish was restricted to 13 out of our 74 participants. These findings are consistent with Dos Santos and Cunha Silva (2005), who reported that Brazilian students of Spanish were aware of a-marking in Spanish yet also found that it represented an acquisitional challenge.

Finally, following Yokota’s (2001) consideration that BP-speaking L2 learners of Spanish could reinterpret a-marking as a prepositional verb regime, we explored individual results in order to find participants who marked objects preceded by the same verb regardless of the animacy features of the object. This analysis was allowed by the fact that we included the same verbs in the two conditions (animate vs. inanimate direct object) under examination. As among the participants whose productive knowledge of DOM was not driven by animacy, we only found three participants who produced a-marking with the same verb in both conditions, we argue that the reinterpretation of a-marking as a prepositional verb regime in L2 Spanish is a restricted acquisitional outcome. Additionally, those three participants only produced this potential reinterpretation of a-marking with four verbs in total.

In sum, the present study documented the acquisition of DOM in Spanish by BP-speaking L2 learners. We found that their acquisition process is consistent with previous hypotheses on the role of L1 transfer, feature reassembly, and the difficulty of functional morphology, particularly in production. Our study shows that BP-speaking L2 learners of Spanish go through an initial L1 transfer stage, followed by a feature reassembly process in which the production of DOM is highly variable. Finally, some of the participants in our study acquired the animacy-driven distribution of DOM in Spanish. Furthermore, this acquisition process is modulated by productive vocabulary knowledge in the second language particularly: the results from the inferential statistics combined with the individual analyses seem to indicate that L1 transfer is present in the L2 learners with the most limited productive vocabulary knowledge while animacy-driven DOM distribution, as per productive knowledge, is featured by L2 learners with deeper productive vocabulary knowledge in Spanish. We take these findings as evidence of an existing correlation between the two variables, and not as the result of causation, as we are aware that an increased L2 proficiency can occur due to numerous factors related to exposure and use. L2 exposure and use patterns may arguably be behind the correlation of the development of these two phenomena: continued opportunities for L2 input and output may lead to the acquisition of both lexicon and syntax, whereas those L2 learners with reduced input and output may not present the same acquisitional outcomes. Finally, the syntactically restricted a-marking in BP may facilitate the identification of the morphology of DOM in Spanish, yet it does not seem to play a significant facilitative role in the acquisition of this phenomenon, unlike among L2 learners whose first languages do have a DOM system. This is not surprising given that BP is described as a non-DOM language.

7. Conclusions

This study documented the acquisition of DOM in L2 Spanish among L1 BP speakers. We found that, after an initial L1 transfer stage, L1 BP speakers start reassembling the animacy features involved in their L2 DOM, which leads to high variability in their productive knowledge of DOM. Finally, a minority of the L2 learners in our study showed animacy-driven knowledge of DOM in their production. Our results indicate that productive vocabulary knowledge can predict the acquisition of the animacy-driven nature of DOM in Spanish among this group of L2 learners, as per their productive and receptive grammatical knowledge, consistently with previous studies reporting that lexical knowledge goes hand in hand with overall proficiency in bilinguals. To our knowledge, the acquisition of this phenomenon is still underexplored in this language pair. Furthermore, this is the first study that uses a productive vocabulary test specifically designed for the language pair consisting of Spanish and BP.

Nevertheless, this study is not without limitations. Our individual analyses show a high degree of variability in the L2 learners, which calls for the need for further research on the role of other variables that may account for these different outcomes. These variables may include experience with formal instruction (both, in terms of the amount and the instructional approach), naturalistic exposure via immersion experiences, and patterns of language exposure and use at home and in the community in Brazil and abroad. Additionally, we acknowledge that forcing participants to provide a reason for every rejection in the AJT (rating a sentence ‘1’ or ‘2’) may have deterred them from choosing these values and selecting ‘3’ instead.

Regarding the validity of the PLEP, we compared the scores of the PLEP and the DELE in Spanish. We found a ceiling effect and a lack of variability in the DELE scores, which suggests that this test might have low predictive power for this population. Overall, these findings suggest that the PLEP can be an appropriate measure to assess overall proficiency in Spanish and BP, whereas the DELE fails to do so given the similarities in this language pair. While we decided to include only cognates in the verbs and direct objects featured in the test items of the tasks in this study in order to prevent confounding effects and to allow all participants to be able to complete the tasks despite their varied proficiency levels, we acknowledge that further research could explore the role of cognates in the acquisition of this phenomenon. This view is consistent with studies on the acquisition of DOM that found that DOM knowledge may be linked to the integration of features in a subset of the lexicon, namely verbs (Hur 2022).

The PLEP is a novel test; thus, supporting evidence for the accuracy of this instrument and for the consistency of its results needs to be gathered from further experimental studies. For that reason, one of the goals of this study (in addition to examining the acquisition of DOM) is to make the PLEP an open-access resource for researchers and educators working with this language pair. We hope that it can contribute to the increasing body of literature on the role of lexical knowledge on different aspects of L2 grammar development and that it facilitates proficiency assessment practices among Spanish-BP bilinguals.