Spanish Clitics and Object Drop

Unstressed object pronouns in Spanish (e.g., lo, la), known as clitic pronouns, are syntactically and phonologically dependent on a verbal host. The focus of this study is on third-person accusative object clitics. Object clitics agree in gender and number with the antecedent to which they refer, as the example in (4) shows.

1. (4) Ramón lo_i compró (el libro).

   Ramon cl-3sg masc bought (the book)

   “Ramon bought it (the book).”

Under certain circumstances, they coappear with the noun phrase (NP, currently labeled “DP” for determiner phrase) to which they refer, or with a null associate (see Borer, Reference Borer1984; Kayne, Reference Kayne1991; Sportiche, Reference Sportiche, Rooryck and Zaring1996). The clitic may be moved or base generated (Cinque, Reference Cinque1990), and it shares phi-features (i.e., gender, number, person, and case) with the fronted DP. Agreement between [+definite +specific] features of the clitic and the DP satisfies the features of the functional head (Uriagereka, Reference Uriagereka1995). Putting aside controversies between movement (Kayne, Reference Kayne1991; Rizzi, Reference Rizzi1986) and base-generation approaches (Everett, Reference Everett1996; Suñer, Reference Suñer1988), it is assumed that clitics behave like agreement morphemes (Borer, Reference Borer1984). To that end, Suñer (Reference Suñer1988) proposes the matching principle, which stipulates that the values of the entry features of clitics always have to agree with those values of the DP antecedent. Under recent views, a clitic does not stand for the null object pronoun but either identifies it or restricts the recoverability of its potential antecedent (Cummins & Roberge, Reference Cummins and Roberge2005; Sigurðsson, Reference Sigurðsson2011).Footnote ¹

Although Spanish is generally not described as an object-drop language, null objects can be used to refer to indefinite antecedents (Campos, Reference Campos1986; Sánchez, Reference Sánchez, Authier, Reed and Bullock1999; Sánchez & Al-Kasey, Reference Sánchez and Al-Kasey1999). This distribution is licensed by the interpretative features of definiteness and specificity— universal semantic categories instantiated in NPs (Farkas, Reference Farkas2002; Fodor & Sag, Reference Fodor and Sag1982; Lyons, Reference Lyons1999). Definiteness refers to whether the NP is known or identifiable by the speaker or hearer (e.g., definite or proper nouns) or not known or identifiable (e.g., indefinite nouns) (Givón, Reference Givón and Greenberg1978). Specificity refers to the referentiality of the NP antecedent (i.e., the existence of the referent in the discourse). Under this view, DPs can be categorically identified as ±specific (Farkas, Reference Farkas, Nash and Tsoulas1995; von Heusinger, Reference von Heusinger2002). Definite DPs are generally specific but can also take a nonspecific interpretation. Indefinite nouns, in contrast, are usually nonspecific but can also have a specific interpretation where differential object marking encodes this distinction (Aissen, Reference Aissen2003).

Definite objects—specific or nonspecific—and indefinite specific objects cannot be dropped in Spanish (Campos, Reference Campos1986; Sánchez, Reference Sánchez, Authier, Reed and Bullock1999). However, indefinite, nonspecific objects must be dropped, as (3b) showed (e.g., Sí, compré Ø; “Yes, I bought [it]”) (Campos, Reference Campos1986). These interpretative and syntactic features also apply to contexts in which the DO appears preposed in sentence-initial position in topicalization or clitic left dislocation (CLLD) constructions (Cinque, Reference Cinque1990; Rivero, Reference Rivero1980; Zubizarreta, Reference Zubizarreta1998). Traditional syntactic analyses in Spanish and Italian have treated preposed topics in these constructions as different types of CLLD (Cinque, Reference Cinque1990; Zagona, Reference Zagona2002). In a topicalization structure, the topic is a base-generated, nonspecific object DP in clause-initial position marked by a specific intonation pattern. This structure does not have a clitic. The nonspecific DP (topic) and its comment (i.e., what is said about the topic) are thought to be linked by a null anaphoric operator as shown in (5a). In the CLLD construction, a base-generated definite topic appears in sentence-initial position. This topic must be doubled by a clitic as in (5b).

1. (5) a. Bare DP +topicalization:

   Café, (*lo) tomo todas la mañanas.

   Coffee, (*cl) I-drink every the mornings

   “Coffee, I drink every morning.”

   b. Definite DP +CLLD:

   Estos zapatos, los compré la semana pasada.

   These shoes, them I-bought the week last

   “These shoes, I bought last week.”

The null object in (5a) or the use of a clitic in (5b) is constrained by the specificity and definiteness features of the topic DP. This bundle of formal features [±definite, ± specific] is not available in Chinese, and this absence may condition a potential role for crosslinguistic influence (Zapata, Sánchez, & Toribio, Reference Zapata, Sánchez and Toribio2005) from Chinese into Spanish.

In current Minimalist analyses, topicalization and CLLD contexts are not distinct constructs. Clitics are a morphological reflex of the silent object; they realize agreement features. Clitics themselves do not license null objects but serve to constrain their recoverability (Cummins & Roberge Reference Cummins and Roberge2005; Sigurðsson, Reference Sigurðsson2011). For instance, a masculine clitic restricts potential linkers to masculine DPs; all overt clitics restrict the selection to definite antecedents. Overt sentential topics—as in (5)—are merged at the left periphery of the clause and trigger an agreement relation with the silent object in argument position. Whereas topics with [+definite, +specific] features match the agreement features of the clitic, indefinites do not. In both cases, the overt topic constrains the identification of the definite or indefinite argument. This is also the case for discourse-introduced topics; thus, note the parallels between (6a) and (6b):

1. (6) a. Indefinite argument drop:

   ¿Tomaste café? Si, ya tomé Ø.

   You-drunk coffee? Yes, already I-drunk

   “Did you drink coffee? Yes, I already did.”

   b. Definite specific topic (DO clitic):

   ¿Qué hace María con los zapatos? Se los pone.

   What does Maria with the shoes Self cl-3pl-masc put

   “What is Maria doing with the shoes? She is trying them on.”

Object Omission in Chinese

Chinese has a restricted pronoun system and does not encode for definiteness morphologically. Third-person singular object pronouns are all marked by the morphological tonic form tā “him, her, it.” This is the same form used for third-person singular subject pronouns. The plural is formed by adding the suffix -men to the pronoun (i.e., tāmen “they”), and there are no gender or animacy distinctions in second- or third-person pronouns (Li & Thompson, Reference Li and Thompson1981). Additionally, Chinese allows null objects regardless of the definiteness or specificity of the antecedent. According to Huang (Reference Huang1984), in topicalization cases null objects can be a variable bounded by a topic in the discourse. That is, they are bound not by a nominal argument in the matrix clause (i.e., A-bound) but rather by the preceding topic in the discourse, as in (7):

1. (7) Zhangsan _ishuo Lisi bu renshi e _*i.

   Zhangsan say Lisi not know

   “Zhangsan_i said that Lisi did not know (him _*i).” (Huang, Reference Huang1984, p. 53)

The null object in (7) cannot be referentially dependent on the subject in the main clause (i.e., Zhangsan) because of intervention effects. It must therefore be referential with someone else outside the sentence—a topic—and treated as a variable (Huang, Reference Huang1984). This means the null object is c-commanded by a discourse-level topic operator, just as in cases in which the topic is present in addition to a nominal argument in the matrix clause. See (8), from Huang (Reference Huang1984):

1. (8) Neige ren_i, Zhangsan shuo Lisi bu renshi e_i.

   That man, Zhangsan say Lisi not know

   “That man, Zhangsan said Lisi didn’t know.”

In (8), the null object is bound by the topicalized NP (i.e., neige ren “that man”) and not by the sentential subject NP. Huang argues that Chinese is a discourse-oriented rather than a sentence-oriented language. This means that topics can be null given that they can be identified with a topic in a topic chain (Huang, Reference Huang1984; Yuan, Reference Yuan1997). However, there is no specification of nominal features of definiteness or specificity and no requirement as to what type of topic can license a null object.

Sigurðsson (Reference Sigurðsson2011) suggests that null arguments are the result of distant agreement with a silent but syntactically present left periphery topic via an invisible C-edge linker (i.e., discourse-linked null topic).Footnote ² He proposes that in languages like Chinese, which lack lexical complementizers, there are no movement constraints on these relations, unlike in the verb-second Germanic languages. Furthermore, there are no feature constraints on the interpretation of the null argument such as would be imposed in object agreement or clitic languages such as Spanish. Table 1 presents a summary of the differences between Spanish and Chinese.

Table 1.

Summary of syntactic and semantic restrictions on object drop in Spanish and Chinese

The Task of the Bilingual Learner

Given the differences shown in Table 1, Chinese-speaking learners of Spanish are exposed to a structure with different potential representations. Spanish null objects are allowed but restricted to indefinite nonspecific referents. The learners’ task is, in part, to acquire the syntax and morphophonology of the clitics. A second dimension of this task is to restrict learners’ L1 object-drop options; they must acquire sensitivity to the definiteness and specificity features involved in clitic agreement in Spanish. This leaves open the possibility of semantic transfer from the L1 (Faure & Bruhn de Garavito, Reference Faure and Bruhn de Garavito2010; Yuan, Reference Yuan1997).

The goal of the present study is to consider the three dimensions in the acquisition process—that is, structure, experience, and influence of the L1. In considering the structural dimension, the potential learnability issues that arise from the acquisition of clitics in Spanish among Chinese-Spanish bilinguals were identified. Thus, a dissociation of morphosyntactic from morphosemantic learning is proposed. In considering experience as a dimension in the learning process, two separate factors are confounded: (a) quantity of input in the learner’s bilingual environment and (b) age of onset of bilingualism. In immigrant populations, age of arrival determines quantity of input in both L1 and L2 (Jia, Reference Jia1998). Age of onset of bilingualism yields three qualitatively distinct types of bilinguals. Most researchers agree that simultaneous bilingual children (i.e., those exposed to two languages before the age of 3;0) can develop functionally differentiated or autonomous systems with little or no systematic interdependence between the languages (De Houwer, Reference De Houwer1990; Genesee, Reference Genesee1989; Paradis & Genesee, Reference Paradis and Genesee1996). Some authors identify some degree of cross-language influence in simultaneous bilingual acquisition (Müller & Hulk, 2001; Yip & Matthews, Reference Yip and Matthews2007), but the interpretation of the facts is questioned (Meisel, Reference Meisel2007). Early child L2 learners—defined in terms of exposure to the L2 after age 3 but before 7—often show less cross-language influence than adult L2 learners (Gavruseva, Reference Gavruseva, Greenhill, Hughes, Littlefield and Walsh1998; Haznedar, Reference Haznedar, Hughes, Hughes and Greenhill1997; Montrul, Reference Montrul2008; Unsworth, Reference Unsworth, Gavruseva and Haznedar2008).Footnote ³ Adult L2 learning is characterized by pervasive patterns of transfer from their L1 (Hawkins & Chan, Reference Hawkins and Chan1997; Jarvis & Odlin, Reference Jarvis and Odlin2000; Liceras, Reference Liceras, Gass and Schachter1989; Montrul, Reference Montrul2000; White, Reference White1987).

For some authors, transfer fundamentally determines the nature of the L2 grammar. For instance, proponents of the failed functional features hypothesis have argued that L2 learners with typologically different L1s are destined for nonconvergence and representational deficits mostly due to maturational reasons or lack of recovery from L1 options (Hawkins & Chan, Reference Hawkins and Chan1997; Hawkins & Franceschina, Reference Hawkins, Franceschina, Prévost and Paradis2004; Tsimpli & Roussou, Reference Tsimpli and Roussou1991). This view has been contested by other researchers who argue for nativelike attainment among adult learners and access to principles of Universal Grammar (Bruhn de Garavito & White, Reference Bruhn de Garavito, White, Pérez-Leroux and Liceras2002; Schwartz & Sprouse, Reference Schwartz and Sprouse1996; White & Genesee, Reference White and Genesee1996). The present study aims to investigate these three dimensions of the acquisition process by (a) looking at the semantic restrictions imposed on the null object and clitic alternation in Chinese-Spanish bilinguals, (b) studying populations with different ages at the onset of bilingual acquisition, and (c) observing differences between bilinguals from birth and late L2 acquirers in the potential influence of Chinese on Spanish.

Studies on the Acquisition of Spanish Clitics

One area in which L2 learners’ ability to instantiate new functional categories has been studied is clitics in Romance languages (Herschensohn, Reference Herschensohn, Prévost and Paradis2004; Sánchez, Reference Sánchez2003; White, Reference White and Clahsen1996). In the case of Spanish, several studies—with English as the L1—have provided evidence of ungrammatical object drop or DP overgeneration (Sánchez & Al-Kasey, Reference Sánchez and Al-Kasey1999; Valenzuela, Reference Valenzuela, Brugos, Clark-Cotton and Ha2005). Other studies at intermediate and advanced levels rule out L1 transfer effects (Bruhn de Garavito & Guijarro-Fuentes, Reference Bruhn de Garavito, Guijarro-Fuentes, Costa and Freitas2002; Zyzik, 2008).

Arche and Domínguez (2011) examined the morphological development of object clitics in L2 Spanish by L1 English speakers under the assumption that clitic-DP matching implies the acquisition of new morphology rather than a new syntactic feature. A question-and-answer task showed a low overgeneration of null objects by beginner and intermediate learners while favoring the production of overt DPs. The advanced learners used clitics productively, which suggests that the acquisition of clitics is possible. Results from a picture-selection task (i.e., comprehension task) showed increased accuracy rates. The level of target responses increased with level of proficiency, especially with morphologically transparent nouns. The authors concluded that learners have an unimpaired narrow syntax despite apparent inflectional variability.

Bruhn de Garavito & Guijarro-Fuentes (Reference Bruhn de Garavito, Guijarro-Fuentes, Costa and Freitas2002) also documented no difficulties with clitic generation in Spanish. The authors examined the role of input in the acquisition of the subjacency constraints associated with object drop in L2 Spanish by L1 speakers of English and Brazilian Portuguese (BP). They tested whether English and BP intermediate learners have knowledge that object drop is ungrammatical with definite antecedents. Brazilian Portuguese allows object drop in most contexts, which can lead to potential null object overextension in Spanish L2. Thus, they predicted clitic overextension among the L1 English speakers in places where a null object is required in Spanish. The results of a grammaticality judgment task (GJT) showed (a) a high degree of acceptance of object clitics with definite antecedents, (b) target gender agreement between clitic and antecedent, and (c) a high degree of acceptance of indefinite object drop. The authors concluded that both groups have knowledge of object-drop constraints in Spanish and make distinctions that would not be possible if L2 acquisition was fundamentally different from L1 acquisition. However, it should be noted that their results could have been confounded by the fact that specificity was not isolated from definiteness in their materials.

In another study, Zyzik (2008) tested the acquisition of third-person object clitics in anaphoric contexts under coordination and under short question-answer dialogues among English-speaking learners of Spanish. Results showed a high production of clitics. The few null objects that were in the data had nonhuman referents and appeared in pragmatically appropriate contexts. The L2 speakers had minimal acceptance of ungrammatical null objects, and this acceptance decreased with higher levels of proficiency. Learners who produced few clitics were more likely to accept null objects in the GJT. Zyzik thus argued for incorrect parsing strategies rather than grammatical errors.

Other researchers have identified difficulties with object drop. Sánchez and Al-Kasey (Reference Sánchez and Al-Kasey1999) found partial specification of the relevant features. Their English-speaking learners of Spanish showed evidence of clitic use but failed to produce and recognize clitics as licensers of CLLD structures or to distinguish the phi-features relevant to clitic doubling constructions. In the production task, the occurrence of null objects was minimal, but the participants showed a preference for the production of overt DPs when clitics were preferable. The authors thus proposed that the learners’ task is to acquire the relevant feature specifications (i.e., definiteness, specificity, and animacy) for the object agreement head (AgrO). The specification of these feature values accounts for clitic drop in certain contexts (e.g., indefinite, nonspecific settings) but not in others (e.g., definite, specific settings).

Valenzuela (Reference Valenzuela, Brugos, Clark-Cotton and Ha2005) also examined the acquisition of the specificity restrictions involved in Spanish CLLD constructions among L1 English speakers. Results showed target levels of clitic use in L2 Spanish with root and embedded specific objects in the sentence-selection task. The participants, however, incorrectly selected the clitic option with nonspecific topics, especially with embedded clauses. In a sentence-completion task, Valenzuela found target clitic use with definite topics, but nontarget use in nonspecific contexts. The speakers seemed to acquire the structure, but they did not seem to be aware of the specificity constraints. The author concluded that abandoning an L1 property is more difficult than acquiring a new L2 property.

More recently, Borgonovo and colleagues (2009) examined the syntactic correlates of specificity among BP learners of L2 Spanish. Specifically, they investigated whether these learners could acquire the constraints of specific objects and whether determiner definiteness could override specificity constraints. In spoken BP, object drop is constrained by the pragmatic and semantic properties of animacy and specificity (Cyrino & Lopes, Reference Cyrino, Lopes, Kepser and Reis2005). Specific or nonspecific targets with inanimate referents must be dropped, whereas animate specific antecedents are overt, and animate nonspecific objects may be optionally realized. They found that L2 learners showed no significant differences from the controls in specific contexts. In nonspecific contexts, however, the L2 learners showed a preference for clitics and accepted gaps less often than did the control participants. Similar results were found with definite and indefinite antecedents. Borgonovo and colleagues argued that overt clitics are underspecified in the L2 grammar with respect to the feature [+specific], which allows them to appear in [+specific] and [−specific] contexts. In contrast, null operators—related to the gap option—are fully specified for [−specific] because they are restricted to contexts with a nonspecific antecedent DP. Thus, the L2 learners accepted an overt clitic in most contexts and tended to reject gaps in all contexts. Difficulties (i.e., overgeneration of overt clitics) seemed unidirectional, a bias that the authors did not associate with L1 effects. Borgonovo and colleagues argued that these patterns of overextension go beyond L1 distinctions and therefore do not result from transfer. However, there is a learnability issue to be considered, which arises for BP learners when null objects are constrained not only by specificity but also by animacy features—a pragmatic constraint that the authors did not control for in their study. Animacy features are not relevant to pronominal object expression in Spanish. Thus, in the absence of sufficient data on the relevant structure, the learners might have overgeneralized the Spanish clitics to all contexts, even if that option is not available in their L1.

Grüter and Crago (Reference Grüter, Crago, Franich, Iserman and Keil2010) examined the role of transfer in the acquisition of L2 French clitics among L1 Chinese and Spanish children. The authors predicted fewer difficulties for Spanish natives and more overacceptance of ungrammatical null objects among the Chinese learners. An elicited production task supported the predictions; however, the results of a truth value judgment task (TVJT) showed no transfer of referential null object in French. It is possible that the results might be due to the methodology employed, and Chinese speakers might simply have shown a preference for the intransitive reading. Costa and Lobo’s (Reference Costa, Lobo, Baauw, Drijkoningen and Pinto2007) replication with the same methodology in European Portuguese showed speakers treated the objectless sentence not as ambiguous but as exclusively intransitive.

Even though these studies have focused on the acquisition of clitics, they have implications for the status of null objects. Sánchez and Al-Kasey’s (Reference Sánchez and Al-Kasey1999) study considers definiteness and specificity features, such as the licensing conditions necessary for object drop. Borgonovo and colleagues’ (2009) study also makes a direct contribution to the study of null objects by investigating the syntactic correlates related to the semantic notion of specificity and sensitivity to null objects in indefinite, nonspecific contexts. Moreover, Yuan (Reference Yuan1997) examined the role of transfer and reduced L2 input in the acquisition of null subject and object constraints by Chinese-speaking learners of English. He found target results in the acquisition of null subject constraints, but the participants were unable to detect the ungrammaticality of null objects in English, even at high levels of L2 proficiency. Yuan suggested that the unlearning of null subjects is triggered by positive evidence in the input, and difficulties with ungrammatical object drop were argued to stem from L1 transfer effects. The [+topic drop] option is retained by the learners, and the relevant input for resetting this parameter is not sufficient in the experience.

Hypotheses

In the previous subsections, a domain of bilingual acquisition in which the L1 represents the least constrained instance of null objects was identified. Because (a) there are no featural or syntactic restrictions on the recoverability of silent arguments in ChineseFootnote ⁴ and (b) null arguments are exclusively discourse linked with no definiteness or specificity restrictions, Chinese learners of Spanish must acquire the overt syntax of clitics and their more complex (i.e., marked) semantic distribution (i.e., definiteness and specificity constraints). Because the L1 represents the least marked and lesser restricted language, transfer is predicted to be favored on the basis of structural considerations. Acquisition difficulties are expected to be localized in the identification of the semantic distribution of clitics but not in the realization of the clitic itself. The possibility of clitic realization is predicted to be universally available as a basic component of the faculty of language (Sigurðsson, Reference Sigurðsson2011). Learning difficulties may arise in the identification of the featural content associated with either the dropped object or the clitic (Cummins & Roberge, Reference Cummins and Roberge2005). This identification process is constrained by peripheral components of language, including experience and language-specific data-processing principles—namely, context scanning in Sigurðsson’s (Reference Sigurðsson2011) proposal—understood as the identification of the most (structurally or semantically) prominent referential antecedent in context.

In terms of experience, the target population is lifelong immigrants living in a Spanish-speaking context. It is therefore assumed that they would have sufficient input, which should favor attainment of the target property. However, speakers who had access to Spanish at different points in their development were also considered. If L1 transfer plays a role in the L2, Chinese-Spanish bilinguals are expected to show high levels of acceptability and production of ungrammatical null objects in definite, nonspecific contexts (9a) and in definite, specific contexts (9b). It is also expected that L2 learners will not overgeneralize overt clitics in indefinite, nonspecific contexts (9c) in which a null object is required.

1. (9) a. *Luisito nunca se come los vegetales pero (la leche) siempre Ø toma.

   Luisito never self eats the vegetables but (the milk) always Ø he-drinks

   “Luisito never eats his vegetables but he always drinks his milk.”

   b. *Cristina llevó la cartera a la discoteca, pero (el pasaporte) Ø dejó en el hotel.

   Christina took the purse to the discotheque, but (the passport) she-left in the hotel

   “Christina brought her purse to the discotheque but she left her passport at the hotel.”

   c. Mi madre quiere que me case pero yo (compromiso) no Ø quiero.

   My mother wants that myself marry but I (commitment) not want

   “My mother wants me to get married but I don’t want any commitment.”

It is also possible that transfer or attainment operate differently for the three bilingual populations, according to their age of onset of exposure. To this end, an experiment was designed to test the following hypotheses:

1. 1. Child L2 and adult L2 learners will interpret null objects in Spanish as referential (i.e., linked to a discourse antecedent), a property not instantiated in Spanish.

2. 2. Child L2 and adult L2 learners will show significant difficulties in their target realization of overt clitics in anaphoric contexts due to transfer from Chinese options (+DP, +Null).

3. 3. Child L2 and adult L2 learners will show significantly higher levels of production and acceptability of nontarget object omission contexts in which an overt clitic is necessary (i.e., with definite topics). They will not, however, have difficulties with indefinite contexts in which a null object is required in both languages.

4. 4. Simultaneous bilingual speakers will behave as native speakers and show target interpretation of the nonreferential meaning of null objects along with target realization of clitics in anaphoric contexts. They will also reject the production or acceptability of ungrammatical object omission in definite contexts. These participants should behave in a nativelike manner given that they were born and raised in a Spanish-dominant environment with Chinese as a home (heritage) language.

In general, the late learners’ difficulties are predicted to occur despite intense and prolonged exposure to Spanish. To acquire the Spanish semantic specifications, they would be required to block the referential meaning of null objects that continues to be active in their L1 Chinese. Thus, a semantic overextension of null objects in CLLD construction is expected. Such learners may also collapse the distinction between nonspecific, indefinite topics and fronted elements in CLLDs, which would lead to optionality for clitics. Similarly, it is predicted that the child and adult immigrants will show a preference for the use of a null object or an overt DP in discourse anaphoric contexts in which a clitic is preferred in Spanish (e.g., ¿Qué hizo Juan con la manzana? Se la comió “What did John do with the apple? John ate it”). For these contexts, as Chinese allows NPs to serve as definites, the L1 is also capable of reinforcing a possible but not preferred option in discourse anaphora in Spanish (e.g., ¿Qué hizo Juan con la manzana? Se comió la manzana “What did John do with the apple? John ate the apple”).

Participants

A total of 53 participants took part in the study. There was one control group (n = 15) of monolingual speakers of Peruvian Spanish and three experimental groups of (a) simultaneous bilinguals (n = 12) who were born and raised in Peru, (b) childhood immigrants (n = 13; age of immigration: 4–14 years old), and (c) adult immigrants (n = 13; age of immigration: 15+ years old).Footnote ⁵ The age of 15 was chosen as the cutoff point for adult bilingualism following previous research that documented significant long-term effects in L2 acquisition outcomes after puberty (see Coppieters, Reference Coppieters1987; Jia & Aaronson, Reference Jia, Aaronson, Greenhill, Littlefield and Tano1999; Johnson & Newport, Reference Johnson and Newport1989). All the participants were recruited and tested in Lima, Peru. They completed a language history questionnaire and a Spanish proficiency test. The proficiency test consisted of an adapted vocabulary section from a Modern Language Association language test and a cloze test, which included three multiple-choice options for each blank. The passage for the cloze test was adapted from the Peruvian newspaper El Comercio. Following previous research, scores between 40 and 50 points were considered the baseline for advanced proficiency level, scores between 30 and 39 points were considered the baseline for intermediate proficiency, and scores between 0 and 29 points were considered as low proficiency (see Montrul & Slabakova, Reference Montrul and Slabakova2003).

The control group consisted of participants who were all undergraduate students from the Pontificia Universidad Católica del Perú. Their average age at time of testing was 21 years old, and their average mean on the proficiency test was 48 out of 50. None of the participants had knowledge of Quechua or any other native language of Peru.

The simultaneous bilingual group consisted of 12 Chinese heritage speakers born and raised in Peru. Their average age at time of testing was 28 years old, with 83% being university educated or currently attending university and 17% having completed high school. Only 7 of the participants reported receiving formal instruction in Chinese in either primary or high school (i.e., the Colegio Peruano Chino 10 de Octubre, a private Chinese school in Lima). The remaining 5 did not report receiving formal instruction in Chinese. Their main language of instruction at the university was Spanish. In terms of language exposure and use at school, 67% reported speaking only Spanish, and 33% reported speaking mostly Spanish. At home, 33% reported speaking mostly Spanish or a bit more Spanish, whereas 17% reported speaking only Chinese, and 50% reported speaking mostly Chinese or a bit more Chinese. At work and in social situations, the majority of the participants indicated speaking Spanish only or mostly Spanish. Finally, 83% of the participants reported feeling more comfortable speaking Spanish. Their average mean in the Spanish proficiency test was 47 out of 50 points. An independent-samples t test showed no significant differences compared to the controls, t(25) = 1.40, p = .171.

The childhood immigrant group consisted of 13 Chinese-speaking immigrants to Peru with an average age at time of testing of 24 years old. Their average age at immigration was 10 years old (range = 4;0–14;0), and their average length of residence in Peru was 14 years (range = 7–22). Most participants (77%) were university educated or were currently enrolled in university, and the rest had completed high school. Nine participants received instruction in Chinese during primary school in China, and 7 received Chinese instruction at the Colegio Chino in Lima. Only 1 participant from this group did not report receiving formal instruction in Chinese in primary or high school. Eighty-five percent of subjects indicated speaking Spanish only or mostly Spanish at school, and 77% reported speaking Chinese only or mostly Chinese at home. In social situations, 54% reported speaking both languages, and 46% reported speaking only Spanish or mostly Spanish. Their average mean on the Spanish proficiency test was 43 out of 50 points (e.g., advanced proficiency). An independent-samples t test showed significant differences when compared to the control participants’ proficiency level, t(26) = 2.57, p < .016.

The adult immigrant group consisted of 13 individuals who immigrated to Peru between the ages of 15 and 30. The mean age of onset of exposure to Spanish was 22 years old, and the mean length of residence in Peru was 10 years. Participants’ average age at time of testing was 31 years old (range = 22–37). All of them received some instruction in Chinese during either primary school or high school. Sixty-nine percent of the participants were university educated, and 23% had completed high school. Sixty-nine percent spoke Chinese only or mostly Chinese at home. At work, 62% reported speaking mostly Spanish or a bit more Spanish, whereas 38% reported speaking both. In social situations, 54% participants reported speaking mostly Spanish or a bit more Spanish, whereas 31% indicated speaking both. About half of them said they felt more comfortable speaking in Chinese. Their average mean in the Spanish proficiency test was 32 out of 50 points (e.g., intermediate level), which was significantly different from the controls, t(26) = 5.792, p < .000.

Structures under Analysis

To investigate whether advanced Chinese-Spanish bilinguals have knowledge of the interpretative constraints of object drop in Spanish, five interpretative conditions were examined in which Spanish and Chinese diverge: (a) the non-coreferential interpretation of Spanish null objects in negation contexts (clitic required), (b) definite specific objects in anaphoric contexts (clitic or DP required), (c) definite nonspecific objects in left-dislocated contexts (clitic required), (d) definite specific objects in left-dislocated contexts (clitic required), and (e) left-dislocated indefinite nonspecific contexts (clitic is not required in either language). Table 2 presents a summary of the structures tested and the tasks used to elicit them.

Table 2.

Summary of conditions examined

Tasks

A total of four tasks were implemented: (a) a question-after-story task (i.e., clitic elicitation task), (b) a TVJT using discourse anaphora, (c) a sentence-completion task, and (d) an acceptability judgment task (AJT). The tasks were conducted orally with the aid of PowerPoint. The preambles and test items were digitally recorded by a native speaker of Peruvian Spanish from Lima. The speaker was a male university student of a similar age to the participants. The participants were told that this was a language study, and all the tasks were conducted in one session.