The Acquisition of Copula Alternation Ser/Estar and Adjective in L1 Russian, Spanish Heritage Speakers

Abstract

Spanish copula choice ser/estar and the semantic and pragmatic distinctions that derive from their alternation in predicate adjective constructions have been discussed in several studies focused on the features of Spanish as a heritage language, usually focusing on the lack of equivalence between English and Spanish. The aim of this study is to determine the competence of a group of heritage speakers of Spanish that were born and raised in Russia in adjective copula selection for ser and estar and to what extent it differs from that of L2 speakers. A group of second-generation heritage Spanish-Russian speakers (n = 29) and a group of L1 Russian learners of Spanish as foreign language (n = 23) performed a translation recognition task in Spanish based on extracts from contemporary Spanish literary works. From a crosslinguistic perspective, a partial correspondence can be established between long forms of the Russian adjective with ser, and short forms of the Russian adjective with estar. Taking this cross-language relationship into account, we considered congruent and non-congruent cross-language scenarios. The results confirm that the heritage speakers outperformed the L2 Spanish speakers. This suggests a possible benefit of earlier exposure and use of Spanish. The facilitative effect of L1 can be traced in the ser-preferred scenarios but it fades away in the estar-preferred contexts for both groups.

1. Introduction

Research on heritage languages (HL henceforth) has witnessed rapid growth in the last two decades (Montrul 2023 and references therein). A growing number of studies has focused on the distinctive features of Spanish as a heritage language (Pascual y Cabo 2016 among others), namely tempo-aspectual relationships (e.g., Cuza et al. 2013) or the double copula of the Spanish verbs ser and estar (Eng. ‘to be’; Cuza et al. 2021; Requena and Dracos 2021; Silva-Corvalán 2014). The semantic and pragmatic distinctions that derive from the ser/estar alternation in predicate adjective constructions are usually attributed to aspectual differences (Arche 2012; Fernández-Leborans 1995; Leonetti et al. 2015): estar usually co-occurs with bounded interpretations of the situation and with discourse situation restrictions (Maienborn 2005), which are equivalent to the perfective aspectual perspective of the situation. On the other hand, ser is devoid of temporal content and therefore it refers to an imperfective aspectual reading. Scholars have normally focused on the lack of equivalence between English and Spanish, but other language combinations still need to be studied in detail. This study considers the competence of a group of heritage speakers of Spanish that were born and raised in Russia in the adjective copula selection for ser and estar. In the case of Russian, there is also a set of semantic properties (Kotov 2014) that restricts the choice but, in this case, they are concerned with the choice between long forms (LF) and short forms (SF) of the adjective in predicative constructions. This study explores the impact that the LF/SF alternation may have on the selection of the ser/estar copula in Spanish.

Traditional proposals for the analysis of the development of heritage languages evolved from the assumption of incomplete acquisition (Montrul 2023; Polinsky 2018). From this perspective, it is understood that early exposure to the dominant social language may hinder the complete acquisition of the elements that make up the structure of the HL, that is, speakers’ grammatical repertoire. Overall, this approach focuses on describing these speakers’ ultimate attainment. Putnam and Sánchez’s (2013) proposal, on the other hand, emphasized a developmental description of the acquisition of the HL. Even if they acknowledged the role of input, that is, the underexposure to the L1 and the overexposure to the L2, they are reluctant to endorse it as the driving factor for the development in the HL. On the contrary, the fluctuation of the levels of lexical activation and the strength of the association between functional, semantic, and phonological features seem to be key in the process (Putnam and Sánchez 2013, p. 488). Heritage speakers (HS henceforth) may restructure their L1 based on the feature mapping of the dominant language to which they are exposed from very early on. This explanation rests on the proposal of the Feature Reassembly Hypothesis (Lardiere 2009), according to which the main difficulty in the acquisition of an L2 lies in the mapping of semantic and syntactic features into lexical items. This study focuses on the role of L1/dominant language transfer in the Spanish copula choice in L2 and HS. We proposed a task related to feature opposition restructuring that may be particularly difficult for heritage speakers (Polinsky 2018) that consisted in detecting, selecting, and reassembling the appropriate semantic features of the predicative construction ser/estar + adjective.

Although the influence of L1 on L2 acquisition has been widely documented, the role of dominant language in HL development has not been considered in detail yet (Romano 2021). According to Montrul and Ionin (2010), phenomena in the syntax–semantics interface are as affected by transfer as other syntax–discourse properties in HS. Since the Spanish copula choice is an interface phenomenon, a potential effect of the dominant language could be expected. Aside from cross-linguistic influence, this study also explores the incidence of lexical frequency, following the proposal of individual patterns of language activation and use provided by Putnam and Sánchez (2013) and confirmed by Thane (2023). In sum, this study represents an empirical contribution to the description of the linguistic competence of Spanish heritage speakers in an underrepresented population such that of Russian–Spanish bilinguals.

2. Copula Choice in Spanish and Predicate Adjectives in Russian

Linking verbs mainly define notions such as the existence, identity, or features of the subject to which they refer. The copula is one of the areas of grammar with the highest degree of typological variation (Arche et al. 2019). This section gives an account of the grammatical resources available to Spanish and Russian to construct copulative sentences.

In Spanish, the existence of two major linking verbs (ser and estar) has generated extensive theoretical descriptions (see Leonetti et al. 2015; Marín 2004, for a review). The collocation of different adjectives with one or both verbs has been a subject of major research. According to Mesa Alonso et al. (1993), around 80% of adjectives in Spanish can co-appear with ser and estar. A distinction is generally made between adjectives that combine (almost) exclusively with ser, as in (1), adjectives that combine with estar, as in (2), and those adjectives with no combinatory restrictions, as in (3).

(1)	Julia es	/ *está	constante.
	Julia isSER	/ isESTAR	persevering
	‘Julia is usually persevering/Julia is persevering now.’

(2)	Julia *es	/ está	cansada.
	Julia isSER	/ isESTAR	tired
	‘Julia is always tired/Juana is tired now.’

(3)	Julia es	/ está	feliz
	Julia isSER	/ isESTAR	happy
	‘Julia is a happy person/Julia is happy now.’

The selection of a linking verb in (3) is motivated by the intention of the speaker to highlight some semantic features over others. The ser/estar alternation of the structure (copula + adjective) has been described from its aspectual basis (Camacho 1995; Luján 1981; Schmitt and Miller 2007): the predicates introduced by ser denote imperfective states or features of the subject that are not constrained beginning or ending points, while the combinations with estar designate perfective states and, therefore, are transitory insofar as boundaries are established on them. This aspectual distinction also selects specific discursive and contextual interpretations for each verb. According to Maienborn (2005), estar requires that the predication depend on a topical situation. On the other hand, ser does not need a link between the predication and a topical situation. In (3), to understand that Julia is happy in Julia está feliz, we expect a context that accounts for Julia’s state to satisfy the link to the topical situation required by the verb. However, in the case of Julia es feliz, we can understand that Julia is happy without a direct allusion to any specific situation.

An updated version of this aspectual interpretation leads to the currently most widespread distinction. It considers predications with ser as individual-level and those with estar as stage-level (Fernández-Leborans 1995; Leonetti 1994; Silvagni 2015). This classification distinguishes between (i) stage-level (SL, henceforth) states, which are bounded by a concrete period of time and describe a spacio-temporal slice or stage of an individual, and (ii) individual-level (IL) states, which describe individuals without referring to their temporal stages and, therefore, trigger inference of temporal persistence (Arche 2012). The transition from one category to the other, as it is the case in (3), is often carried out through a process of coercion (Escandell-Vidal and Leonetti 2002). This process allows the semantic content of the adjective to be reinterpreted. Additionally, it triggers a conceptual readjustment that enables the combination with ser and estar. Marín (2010) analyzed IL adjectives and established a distinction between nonrelational adjectives, as in (4), which can be combined with estar in certain contexts, and relational adjectives, as in (5), which allow collocation with the same verb but in a much more restrictive way.

(4)	Hoy	estás	tú	muy	valiente.
	today	areESTAR	you	very	brave
	‘Today you are very brave.’

(5)	?Juana está	muy catalana.
	Juana isESTAR	very catalan
	‘Juana is very Catalan.’

The combination of valiente ‘brave’ and hoy ‘today’ in (4) enables a context in which the subject is anchored to a temporal-spatial situation. The explicitness of the construction verb +adjective is then reinforced by the adverb. In (5), the reinterpretation of catalana (’Catalan’) is only possible if the relational nature of the adjective is canceled and the adjective is considered a non-relational adjective. On many occasions, this alternative reading can only be achieved by the integration of pragmatic/sociocultural knowledge. Hence, it is necessary to know the stereotyped social image that considers Catalans as people who care about money to be able to interpret catalana (’Catalan’) in (5) as ‘stingy’. This study focuses on ambivalent adjectives as those presented in (3) and (4). In these cases, the selection of ser and estar depends on intended meaning.

In Russian, adjectives in the predicative position can appear in the long form (LF henceforth) and in the short form (SF henceforth). Morphologically, LF adjectives differ from SF adjectives in that the latter lack case inflection. Concerning syntax, only LF adjectives can appear in an attributive position within the noun phrase. In Spanish, only ser can be combined with nouns. It is generally accepted that, at the plot structure level, SF adjectives behave more like verbal elements, while LF adjectives are considered purely adjectival (Babby 2009; Geist 2010; Karpacheva 1999). In fact, some scholars have consistently pointed out that the LF adjectives appearing in the predicative position should be analyzed as an attributive adjective followed by a null noun1 (Babby 2009; Bailyn 1994; or Pereltsvaig 2000). From this, it follows that the LF adjectives imply a relative interpretation (6a) and the SF ones, an absolute interpretation (6b).

(6a)	Étot	vor	khoroshij.
	this-M-SG.NOM  thief-M.SG.NOM is-Ø  good-M.SG.NOM
	‘This thief is goodLF’.
(6b)	Étot	vor	khorosh.
	this-M-SG.NOM  thief-M.SG.NOM is-Ø  good-M.SG
	‘This thief is goodSF’.

In his description of the behaviour of Russian LF and SF adjectives, Timberlake stated that “the short form indicates that the subject, viewed as a unique individual rather than as a type, manifests the property in potentially variable ways under different circumstances” (Timberlake 1993, p. 863). In (6b), this property is interpreted as being good as a thief; LF adjectives, on the other hand, “signal that the subject, viewed as a type of individual, instantiates an essence, a quality” (Timberlake 1993, p. 862). This property of goodness in (6a) refers to a more abiding value, which in this case would refer to a person that is morally good (Martin and Bikina forthcoming). From this approach, a possible correspondence emerges between LF/SF of Russian adjectives and ser/estar copula choice in Spanish emerges.

Authors like Shvedova (1980), Grashchenkov (2007), or Kotov (2014) described the difference between LF and SF in terms of the stage-level vs. individual-level. SF adjectives describe a property of the individual as a qualitative state of that individual, that is, a property constrained to a specific state of circumstances, which generates a stage-level interpretation. On the other hand, LF adjectives cannot have an eventive interpretation2. Hence, they cannot be restricted to a particular time associated with individual-level predicates. According to Geist (2010), the distinction between ‘property’ and ‘state’ is of paramount importance. As a result, in (7), the SF adjective denotes the quality of being attentive as an eventuality, which renders the combination with a time delimiter like ‘yesterday’ completely legitimate. On the other hand, any temporal-spatial delimitation in a predicative sentence with LF adjectives will present problems of interpretation. The LF adjective in (8) denotes the quality of being attentive as a property of the person.

(7)	Vchera	rebenok	byl	vnimatelen.
	yesterday	child-M.SG.NOM  was-PST.IPFV.M.SG		attentive-M.SG
	‘Yesterday the child was attentiveSF.’

(8)	? Vchera	rebenok	byl	vnimatelnyj
	yesterday child-M.SG.NOM was-PST.IPFV.M.SG  attentive-M.SG.NOM
	‘Yesterday the child was attentiveLF.’

According to this, and as Geist (2019) pointed out, Spanish and Russian have different grammatical mechanisms to express the link of a predication to a specific situation. A partial correspondence can be established between the LF of the Russian adjectives with Spanish copula verb ser, and the SF of the Russian adjectives with Spanish estar (Denisova 2011; Vinogradov and Miloslavsky 1986), as can be seen in (9). However, the interlinguistic correlation is not perfect, as some syntactic contexts condition the use of LF and SF adjectives in Russian (Nesset and Janda 2023) and, at the same time, several combination restrictions of some Spanish adjectives apply. For example, in (10), the Spanish correlate of the Russian adjective ‘udachlivyj_LF/udachliv_SF’ can only be combined with ser.

(9a)	La	vida	Es/está	difíficil
	the-F.SG	life	isSER/ isESTAR	hard
(9b)	zhizn’		trudnaya	/ trudna
	life-F.SG	is-Ø	hard-F.SG.NOMLF	/ hard-F.SGSF
	‘Life is hard.’

(10a)	Él es		afortunado	en el	trabajo
	he isSER	/ *isESTAR	lucky	in the	work
(10b)	On		udachliv		v rabote.
	he-M.SG.NOM		is-Ø lucky-M.SG		in work-F.SG.LOC
	‘He is lucky at work.’

Given that cross-linguistic influence takes place in any multilingual learning/acquisition process (Safa 2018), and transfer has been claimed as one of the most frequent mechanisms and strategies used by speakers (Jarvis 2017), this study explores the extent of this cross-linguistic correlation on the Spanish proficiency of HL and L2 speakers, whose dominant language is Russian.

3. The Acquisition of the Spanish Copula Choice

Research on the acquisition of Spanish copula choice in ser/estar + adjective has confirmed that this is a grammatical phenomenon associated with a complex learning process that constitutes a difficulty for L2 Spanish speakers (Guijarro-Fuentes et al. 2023; Geeslin 2003, 2014; Perpiñan et al. 2019), even those with advanced levels (Guijarro-Fuentes et al. 2023; VanPatten 1987) or heritage speakers (Cuza et al. 2021; Lowther and Lindsey 2005; Requena and Dracos 2021). The results of intervention studies in formal teaching contexts have shown that the effects of explicit instruction on this construction are limited at initial-intermediate levels of competence (Winitz and Sagarna 2007). As explained in the previous Section, context identification is important to select ser or estar and, in many cases, L2 speakers may be unaware of the semantic and pragmatic nuances that are needed to establish those contexts successfully. In the case of HS, when Spanish is not the dominant language since the early stages of development, it could be expected to find differences compared to other monolingual populations (Montrul 2016), as described by other studies on copula selection and other grammatical phenomena (Silva-Corvalán and Montanari 2008; Pascual y Cabo 2016 among others).

Research on the selection of the Spanish copula is scant as far as HS are concerned. Longitudinal studies on children HS based on spontaneous oral production recordings of English-dominant, Spanish–English bilingual children from 1 to 5 years old indicate that copulative constructions develop autonomously and are heavily influenced by the interactions with the adult(s) in the child’s environment (Silva-Corvalán 2014; Silva-Corvalán and Montanari 2008). These studies indicate a certain delay in the acquisition of estar, which is explained by a possible influence from English.

Two recent studies have addressed Spanish copula selection in school-age and adult HS. On the one hand, Requena and Dracos (2021) carried out a study that examined whether there were differences between second-generation HS schooled in the US (n = 50) and first-generation adults (n = 21) in the knowledge of semantic constraints and pragmatics associated with the selection of the copula in predicative constructions. The competence of the participants in both English and Spanish was measured and the children’s language exposure and use outside school was assessed through a questionnaire that was filled in by parents. The study contained two experimental oral production tasks with Spanish adjectives and novel adjectives that recreated situations of temporary changes that allowed the analysis of the acquisition of aspectual dimensions in copula + adjectives construction. The results indicated that first generation adults’ knowledge is similar to that of monolingual speakers. They also signaled a near categorical selection for estar with temporary contexts. The results of the heritage group confirmed a high heterogeneity in the answers and an effect of the competence in Spanish as a significant predictor of the preference for estar in temporary contexts. More specifically, heritage speakers with higher proficiency in Spanish exhibited adult-like copula selection. In contrast, schoolchildren with lower proficiency exhibited little sensitivity to the aspectual dimensions of copula choice. These speakers showed a chance behavior in the production of estar, especially in the task with novel adjectives, where lexical knowledge was controlled.

In a study on the production of ser and estar in children (n = 16) and heritage Spanish-speaking adults (n = 19), Cuza et al. (2021) used an adaptation of a controlled production test used by Schmitt and Miller (2007) to elicit the use of copula with adjectival predicates and in event locatives constructions in both ser-favored and estar-favored contexts. The performance of heritage-speaking children indicated a lack of sensitivity to the ser/estar distinction. The individual data for adult heritage speakers pointed to a pattern of development that is close to the performance of monolingual adults. Considering the results, it was concluded that “it is possible for these structures to be stabilized in the adult grammar” (Cuza et al. 2021, p. 17). Although much mitigated in the case of adults, the authors also identified a difficulty in the use of ser in individual-level predicates in the two groups of HS, which results in an overproduction of estar in ser-favored contexts. According to the authors, the variability shown by the two groups in this task may be associated with patterns of exposure to the language, the use of Spanish, and proficiency level.

The overgeneralization of estar in HS was also reported by Valenzuela et al. (2015) in their study about heritage speakers of Spanish in the US. The study also included a HS group in Canada, whose performance on all tasks matched that of the monolingual control group. This led the researchers to conclude that typological proximity may entail a facilitative effect in the knowledge of this structure and opened the way for studies that may consider the impact of previously acquired languages on this type of speaker.

Several studies in L2 Spanish contexts have covered various profiles of learners at various levels of competence, from studies focused on intermediate and higher levels (Guijarro-Fuentes et al. 2023; Bruhn de Garavito and Valenzuela 2006; Geeslin 2003) to those considering ultimate attainment (Guijarro-Fuentes et al. 2023). Empirical evidence from most of these works has shown that L2 learners of Spanish do not achieve native-like competence in copula selection. This may suggest that there are differences in the ability to apply semantic and pragmatic constraints in non-native grammars (Guijarro-Fuentes et al. 2023). Even when the distinction between ser/estar seems to be well established, L2 speakers seem to have difficulties related to the discrimination of the associated interpretive properties that are needed for the copula selection process (Bruhn de Garavito and Valenzuela 2006).

There seems to be no unanimity on whether the L1 influences the acquisition of copula choice in Spanish L2. On the one hand, Bruhn de Garavito (2009) signaled a facilitating effect of German versus English in the distinction between eventive and stative passive constructions but did not detect an effect of the L1 on copula selection. Similarly, no L1 effect was found by (Guijarro-Fuentes et al. 2023), who concluded that Spanish learners whose L1 does not have a copula choice follow a similar acquisition process.

On the other hand, more recent works such as (Guijarro-Fuentes et al. 2023) showed evidence of a transfer effect. In this study, the acquisition of the construction estar+adjective in the L2 Spanish grammar of French and Portuguese L1 speakers was assessed. Based on data extracted from a learner corpus, the study aimed at delimiting to what extent the aspectual contrast in linking verbs in Portuguese may have a facilitating effect on the correct use of estar in predicative constructions. Results showed that L1 Portuguese learners clearly outperform L1 French learners. The difficulties experienced by L1 French speakers are understood as a consequence of the features assembly task (Lardiere 2009) that these speakers need to face. These findings are especially relevant for the purpose of our work. Given the partial interlinguistic correspondence described in the previous section between Russian and Spanish, this study delves deeper into the Spanish copula selection in different native and non-native populations.

The aim of the study is to analyze the competence of L2 Spanish, L1 Russian speakers and Russian dominant, Spanish heritage speakers in the use of the ser/estar + adjective alternation. Following the results of the studies described above, we focus on two specific phenomena: crosslinguistic influence and differences between populations with different acquisition profiles. Thus, the following research questions are established:

(i)

To what extent does proficiency in the use of Spanish copula choice in predicative constructions differ between Spanish heritage and advanced L2 Spanish speakers?

(ii)

Considering the properties of the dominant language (Russian), is there any interlinguistic transfer effect, based on the LF vs. SF adjectives from Russian, among HL and L2 speakers on the use of Spanish copula choice in predicative constructions?

4. Methodology

4.1. Participants

The study involved 29 Spanish Heritage Speakers (SPAH) and 25 L2 Spanish (L2SPA) students. All participants were between 18 and 35 years old (SPAH: mean = 27.23, SD = 3.51; L2SPA: mean = 22.82, SD = 0.22). Most heritage speakers (n = 23) came from mixed families, that is, one parent was a speaker of Spanish and the other spoke Russian. The rest of the heritage speakers’ parents could both speak Spanish. Regarding the origin of the families, 19 participants indicated that at least one of their parents was from Spain, seven from Nicaragua and three from Cuba. All of them had been born and raised in Russia. None declared to have had any formal education in Spanish, either at a school, university, or language center. Candidates were recruited through contact groups on different social networks.

To assess their level of competence, a personal interview was conducted in which two oral interaction tasks from the Diploma of Spanish as a Foreign Language (DELE) test corresponding to level B2 were integrated. The evaluation of the interviews was carried out by two Spanish teachers in Russia with extensive experience as examiners and interviewers of the DELE tests. Although oral proficiency tests may not function adequately as placement tests for heritage speakers (Valdés 1989), they can be used as indicators of their general oral proficiency (Kagan and Friedman 2003; Swender et al. 2014). After the interview, a simplified version of the language use patterns questionnaire (Oh and Au 2005) was administered. The proportion of use of Spanish declared by the participants varied between 35–60% (mean = 46.53, SD = 8.08). A total of 41% of the participants (12/29) indicated that they felt more comfortable speaking Russian, while the remaining 59% said that they felt equally comfortable in both languages.

The group of L2 Spanish speakers was made up entirely of students from Saint Petersburg State University. They were attaining a degree in Romance Philology or Translation and majoring in Spanish. To participate in the study, they were required to hand in a copy of a DELE B2 certificate that was a maximum of two years old. All confirmed being L1 Russian speakers and indicated more than two years of formal instruction in Spanish (mean = 3.35, SD = 0.51). No participant indicated having completed immersion stays of more than four months in a Spanish-speaking country, so it can be assumed that their contact with Spanish was almost exclusively formal.

4.2. Materials

Both groups completed an offline written inverse translation task (Russian > Spanish). Translation tasks have been used in second language research (Sunderman 2013), especially to study lexical access and sentence processing (Montrul and Foote 2014; Hyun Lim and Christianson 2013), and crosslinguistic influence (Sharpen 2016). Test stimuli consisted of fragments of translations of three contemporary Spanish novels translated into Russian: Caperucita Roja en Manhattan, by Carmen Martín Gaite, Corazón tan blanco, by Javier Marías, and La sombra del viento, by Carlos Ruiz Zafón. Translation passages containing long and short adjective constructions were selected and checked against the original corresponding passages.

The passages in Russian were classified according to whether of interlinguistic correspondence could be expected and according to the verb used in the original in Spanish. Thus, four conditions were set: (a) adjective in the LF corresponding to a construction ser + adjective in Spanish; (b) adjective in SF corresponding to a construction ser + adjective in Spanish; (c) adjective in LF corresponding to a construction estar + adjective; and (d) adjective in SF corresponding to a construction estar + adjective. The complete test (K = 35) was made up of 20 target fragments (five fragments per condition) plus 15 fragments that served as fillers. The list of target fragments that make up the test in included in Appendix A.

(11)	Bol’she ona	o	nem	nichego	ne govorila,	i	bylo	neponyatno,
	more  she  about he-LOC nothing-M.SG.GEN no say-PST.IPFV.F.SG and be-PST.IPFV.N  incomprehensible-N
	krasivyj			on ili net.
	beautiful-M.SG.NOM  he or no
	‘But she never mentioned anything that made it clear if she was pretty or ugly.’

(12)	Vino	bylo	sovsem deshevo	i	dazhe v schet	ne	shlo.
	wine-N.SG.NOM be-PST.IPFV.N totally  cheap-N.SG and even  in bill-M.SG.ACC  no  go-PST.IPFV.N.SG
	‘The wine was so cheap it was not even included in the bill.’

(13)	Tryapochku	pora	propoloskat’,	pryamo chyornaya.
	rag-F.SG.ACC it’s time		rinse-INF	directly black-F.SG.NOM
	‘You should wash the cloth. It is black.’

(14)	V tot	vecher	on	byl	osobenno	vzvolnovan.
	in that-N.SG.ACC evening-N.SG.ACC he-NOM be-PST.IPFV.M. particularly					disturbed-N.SG
	‘That afternoon, he was particularly nervous.’

A total of 32 fragments were initially selected, eight for each of the four conditions. An extra context utterance was added to these fragments so that the participant could place them within a specific situation. Assuming that the selection of ser/estar plus adjective can be guided on many occasions by discursive and pragmatic properties (Camacho 2012), the context was edited in such a way that it functioned as an indicator against the choice of ser and estar. For example, the target fragment (a) was presented with the following pre-context:

(15)	Susana llevaba semanas hablando de su amigo César con su amiga Carmen. Carmen empezaba a sentir un interés
	especial por César. Quería saber un poco más sobre su aspecto físico y le preguntaba insistentemente a Susana.
	Pero nunca mencionaba nada que hiciera entender si era bonito o feo.
	‘Susana had been talking about her friend César with her friend Carmen for weeks. Carmen was beginning to feel a
	special interest in César. She wanted to know a little more about his physical appearance and she insistently asked
	Susana. But she never mentioned anything that could make her understand if he was pretty or ugly.’

To check the validity of the test, the task was administered to ten native Spanish speakers with a high level of proficiency in Russian. All the fragments that reached 85% of coincidence with the expected response, that is, with the copulative verb that appeared in the Spanish original, were kept in the final test version. Alternative translations in the selection of the verb in Spanish were discarded, such as, for example, the translation of vzvolnovan (‘nervous, excited’), which was sometimes translated as se puso nervioso (‘he got nervous’) and in others as estaba nervioso (‘he was nervous’). However, variation in the translation of the adjective was not considered an exclusion factor. For example, Russian adjectives such as krasivij (‘beautiful’) could be translated as guapo and hermoso, according to the variety of Spanish spoken by the native speakers.

The frequency of the Spanish adjectives was controlled. All the adjectives were among the 4000 most frequent words in the lexical frequency ranking of Corpus del Español (Davies 2002). This corpus contains 20,000 words of written and oral texts from Spanish and Latin American texts. None of the selected adjectives had combinatory restrictions with either ser or estar. In this way, it was possible to ease a possible effect of bias in the input on the responses of the participants.

4.3. Procedure and Analysis

The data were collected in Saint Petersburg, between September and December 2018. The test was presented in paper-pencil format and was carried out in the presence of a researcher. The group of HS, whose participants had already been interviewed and had completed the sociolinguistic questionnaire, took the test on the premises of the State University of Saint Petersburg. The L2 learners completed the task during a class session at the same university.

Three versions of the same test were created with the fragments arranged in a different order. The participants were presented with the 35 fragments and were asked to translate only the final part, which was underlined. They were asked not to use reference materials and they were given the opportunity to ask for the meaning of any word in Spanish. The test lasted between 30 and 40 min for both groups.

Data collected from ser-preferred contexts and estar-preferred were analyzed separately. The dependent variable for both analyses was obtained by coding participants’ responses as either ‘1′, when the participant produces the expected copula, and ‘0′ for the unexpected responses—other copula or other lexical verb. The data were analyzed implementing a generalized linear mixed model (GLMM). We applied a binomial probit distribution to fit the binary response. The models generated predicted probabilities that participants would choose ser (for ser-preferred data model) or estar (for estar-preferred data model).

Group (SPAH or L2SPA) and Form of Russian Adjective (long or short) were included in both models as fixed effect factors along with the interaction term. Factors were introduced with treatment code fashion. Participants were included as random intercept varying by Form of Russian Adjective as random slope. Random effects were measured as the adjusted intra-class correlation coefficient and their contribution to the model performance was assessed using the likelihood ratio test, which compares the model with random effects and the model without them. In both cases, low ICCs (ICC_ser-model = 0.14; ICC_estar-model = 0.01) indicated a limited contribution of the random structure to the power of the model. LRT results confirmed this lack of contribution (ser-preferred: X² = 1.046, df = 3, p = 0.790; estar-preferred: X² = 0.053, df = 3, p > 0.90). Considering these results, models without random structure were selected to conduct further analyses.

To evaluate the accuracy of the logistic models AUC-indexes were calculated from the confusion matrices. Results from ser-preferred data model showed a good ROC curve (AUC = 0.709, optimal cutoff = 0.77), and moderate-good classification accuracy (sensitivity = 65%; specificity = 82%). In the estar-preferred data model, the ROC curve reached an acceptable level (AUC = 0.65, optimal cutoff = 0.44), and moderate-good classification accuracy (sensitivity = 83%; specificity = 41%). Both models performed better than chance in predicting the actual participants’ responses. Logistic models were fitted using the glmer and performance functions from lme4 (Bates Douglas et al. 2015) and ROCR (Sing et al. 2005) packages. All analyses were performed using the R software (R Core Team 2020).

5. Results

The recording of the dependent variable as a dichotomous response allows obtaining a descriptive summary of the performance of the two groups of participants.

Table 1. Tokens and percentages of expected production of ser and estar by form of Russian adjectives and group.

	Ser		Estar
	Long Form	Short Form	Long Form	Short Form
SPAH (n = 29)	143 (99%)	112 (77%)	96 (66%)	111 (76%)
L2SPA (n = 25)	114 (91%)	98 (78%)	64 (51%)	55 (44%)

shows the count and proportion of correct responses, understood as expected responses, for the fragments for ser-preferred and estar-preferred depending on the form of the Russian adjective and the type of speaker. In both groups, the proportion of responses expected for ser is higher than that of estar. From the perspective of interlinguistic influence, the response pattern of the SPAH group reproduced the expected behavior in a more clear way in case of influence of the forms of the adjective in the selection of the copula: the scenarios of convergence between languages (i.e., long form-ser and short form-estar) had a higher proportion of expected responses, while the scenarios of divergence between languages (i.e., short form-ser and long form-estar) registered a lower accuracy rate. In the case of L2 learners, the predicted cross-language distribution holds for the ser-preferred fragments, but not for the estar-preferred ones.

The results of the ser-preferred data model confirmed a significant Group effect, whereby belonging to the L2 group implied a decrease in the probabilities of producing the ser form in the translation of adjectives with a long form (b = −1.931, SE = 0.778, p = 0.013). The interaction effect between Group and Adjective Form was also significant (b = 1.998, SE = 0.832, p = 0.016). This suggests that, although, in both cases, the probabilities of selecting ser decreases, the slope for L2 learners from long form to short form is less steep (12.8%) than the slope for HS (21.3%).

Multiple comparative post-hoc tests with Holm correction for p-value adjustment confirmed a significantly higher probability mean of producing the ser form for long adjectives in comparison with short adjectives in the SPAH group (OR = 21.07, z = 4.126, p < 0.001) and L2SPA (OR = 2.86, z = 2.737, p = 0.006). Comparing groups, a significantly higher probability mean was detected for HS in long adjectives (OR = 6.89, z = 2.481, p = 0.013). No significant differences between L2 and HS were found for short adjectives (OR = 0.935, z = −0.228, p = 0.819). As shown in Figure 1, SPAH obtained a higher probability mean of producing ser with long adjectives with a Russian counterpart (0.98, CI95% [0.94; 0.99]) than L2SPA (0.91, CI95% [0.98; 0.95]). Therefore, the interlinguistic convergence scenario seems to favor the production of the expected response in both groups, compared to the contexts of incongruence between languages, which show lower probabilities of producing ser.

For estar-preferred data, the fitted model results showed a significant effect of Group. Again, belonging to the L2 group significantly reduces the probabilities of producing the estar form with long-form Russian adjectives compared to HS (b = −0.624, SE = 0.251, p = 0.012). No significant effect was found for adjective forms (b = 0.511, SE = 0.263, p > 0.05), but a significant interaction effect was confirmed (b = −0.799, SE = 0.365, p = 0.028). As can be seen from Figure 2, while the slope for L2 speakers from long to short form slightly decrees (7.2%), the inverse relation between forms of adjective arises for HS, showing an increasing slope (10.3%) from long to short forms. Multiple comparative post-hoc tests with Holm correction for p-value adjustment confirm a marginally significant higher probability mean of producing estar form for short adjectives in comparison with long adjectives in the SPAH group (OR = 0.60, z = −1.971, p = 0.05). For L2SPA, there were no significant differences (OR = 1.34, z = 1.139, p = 0.254). Comparing groups, a significant higher probability mean was detected for HS in both long adjectives (OR =1.87, z = 2.491, p = 0.012) and short adjectives (OR = 4.16, z = 5.350, p < 0.001).

These results indicate that HS are more prone to use estar than L2 speakers regardless of the form of the Russian adjective. Moreover, the heritage group performed according to the crosslinguistic correspondences. Short forms of Russian adjectives seem to entail a facilitative effect on the preference for estar that is not paralleled in the L2 group. Interestingly, the production of different adjectives by L2 speakers in estar-preferred data was around 50% probability (long form = 0.51, CI95% [0.42; 0.59]; short form = 0.44, CI95% [0.35; 0.52]), which is a behavior very close to a performance by chance. The heritage group, on the contrary, had higher probabilities for both scenarios (see Table 1).

To obtain a more fine-grained picture of the results obtained at the group level, we conducted an individual by-item analysis. For this descriptive analysis, proportions of expected responses were calculated for every context in both groups. On each item, an indicator of frequency symmetry of the combinations in Spanish ser/estar + adjective was also included through the Sketch Engine program (Kilgarriff et al. 2014). This indicator was obtained from the normalized frequency per million words, subtracting the frequency of the combination estar + adjective from that of ser + adjective. Thus, an indicator close to zero represented an almost perfect symmetry between both combinations, while negative values indicated an asymmetry in favor of estar and positive values indicated an asymmetry in favor of ser, which is the expected response. According to this, a large positive value could be associated with a potential positive effect, since the input contained a ser+adjective bias.

Table 2. Percentages of expected production by item of ser in SPAH and L2SPA groups.

	Long Form > Ser					Short Form > Ser
	Item 1	Item 2	Item 3	Item 4	Item 5	Item 6	Item 7	Item 8	Item 9	Item 10
SPAH (n = 29)	100	100	96	96	100	100	93	89	48	55
L2SPA (n = 25)	92	100	80	100	84	100	64	100	36	92
Frequency symmetry	−0.03	1.06	0.08	0.24	-0.30	0.23	0.26	0.00	0.32	0.40

summarizes the results of the ser-preferred context.

The SPAH group showed an at-ceiling performance in most instances with long form adjectives, even when the combination with estar was more frequent, as in the case of feliz (‘happy’) in item 5. Somewhat lower proportions were recorded in the group of L2 learners, although in no case was their precision in the answers less than 80%. The answers of the groups were less stable for the items with the SF adjective. Whereas some contexts registered a high proportion of ser (item 6 and item 8), other items produced precision levels below 50% in both groups, such as item 9 reproduced in (12) and now recovered in (16), or item 10 in Example (17), which was favored by the L2 group.

(16)	Vino	bylo    sovsem deshevo		i	dazhe v	schet	ne	shlo.
	wine-N.SG.NOM be-PST.IPFV.N totally		cheap-N.SG	and	even  in bill-M.SG.ACC		no	go-PST.IPFV.N.SG
	‘The wine was so cheap it was not even included in the bill.’

(17)	Vprochem, ona	eshhe ochen’ moloda	i	so	vremenem	mozhet	izmenitsya.
	however  she-NOM still  very  young-F.SG and with				time-N.SG.INS can-3SG.PRS change-INF.PFV
	‘However, she is still very young and may change over time.’

In (16), being cheap is presented as a property of the wine. However, the eventual interpretation prevails in both groups, which would also be reinforced by the presence of the SF adjective in Russian. The answers obtained from (17) may indicate an influence of the context to which the SPAH is more sensitive and therefore the tendency estar could be reinforced. On the other hand, the L2SPA group mainly selected ser. The SF does not seem to have an influence in the case of L2SPA, nor does the prevalence of the combination ser versus estar with the adjective joven (‘young’).

The results of estar-preferred contexts showed lower expected choice ratios in scenarios of interlinguistic convergence and divergence. For these contexts, the frequencies of ser + adjective were subtracted from the frequencies of estar + adjective. Thus, negative values indicate an asymmetry in favor of ser and positive values an asymmetry in favor of estar, which is the expected response.

Table 3. Percentages of expected production by item of estar in SPAH and L2SPA groups.

	Long Form > Estar					Short Form > Estar
	Item 11	Item 12	Item 13	Item 14	Item 15	Item 16	Item 17	Item 18	Item 19	Item 20
SPAH (n = 29)	31	45	86	79	89	96	100	76	48	62
L2SPA (n = 25)	12	28	76	56	84	56	84	48	12	20
Frequency symmetry	−0.18	−0.15	−0.02	0.16	0.03	0.33	0.69	0.10	−0.05	0.00

shows the results of estar-preferred contexts.

The items that contain SF adjectives recorded a higher frequency of estar in both groups; it is also observed that in the adjectives with estar-bias such as nervioso (‘nervous’) (item 16) and tranquilo (‘quiet’) (item 17), estar was more frequently chosen, especially by the SPAH group. Within this set of items, only the adjective débil (‘weak’) (item 19) did not reach the 50% expected production in any group. The semantic field change implied by the ser/estar alternation in this case could have directed the preference for ser, since it is also a plausible interpretation within the context in which it appears:

(18)	Mariya zabotilas’	o	paciente,	kotoryj	pochti oslep	i
	Mariya take-care-PST.IPFV.F about		patient-M.SG.LOC which-M.SG.NOM almost go-blind-PST.PFV.M and
	byl    ochen’ slab.
	be-PST.IPFV.M  very  weak-M.SG
	‘Maria cared for a patient who was almost blind and very weak’.

The proportions of estar obtained from Russian LF adjectives registered lower values in both groups, even if a statistically significant difference could only be confirmed for the SPAH group. The asymmetry of frequencies shows some correspondence with what could be expected if we assume an effect of the input. Adjectives with a clear ser-bias, such as negro (‘black’) (item 12) or amable (‘friendly’) (item 11), reproduced in (19), account for the lowest productions of estar. The sum of the bias present in the same input and a possible negative transfer effect of the LF adjective seems to have a greater impact in the L2SPA group than among heritage speakers.

(19)	Ona	ochen’	priyatnaya,	tshhatel’no odeta,	kazhetsya	spokojnoj.
	she-NOM  is-Ø very  nice-F.SG.NOM carefully			dressed-F-SG seem-3SG.PRS.IPFV		calm-F.SG.INST
	‘She is very nice, carefully dressed, and seems calm.’

In general, estar-preferred contexts are less transparent than ser-preferred ones. In the search for crosslinguistic influence, this becomes more apparent in the production of estar in heritage speakers. In the individual by-item analysis, the influence of frequency symmetry on the behavior of the L2SPA group stands out.

6. Discussion

Two research questions were formulated in this study. The first focused on the acquisition and development processes of a heritage language versus those of an L2. The results for the production of ser and estar have shown a greater production of expected responses for the ser-preferred contexts compared to the estar-preferred ones in both groups. These results contrast with other similar studies (Lowther and Lindsey 2005; Silva-Corvalán 2001), which found a tendency to overgeneralize the use of ser and a tendency to produce innovative uses of estar. According to our results, the SPAH group outperformed the L2SPA group at least on one condition of ser-preferred contexts, which may indicate that it is possible to stabilize the copulative constructions with ser in the grammar of Spanish HS, as indicated by previous studies (Cuza et al. 2021). In their study, these authors attributed this advantage of adult HS over child HS to the intensive use of Spanish reported by the participants in the administered questionnaire. Their individual analysis allowed them to acknowledge that the adequate domain of ser in these constructions is subject to a dynamic process that depends on the exposure to the HL, its use, and speakers’ level of competence.

Regarding the production of estar, our results are in line with previous studies that identified the estar + adjective combinations as less accessible structures or with a more limited domain, especially for L2 speakers (Bruhn de Garavito and Valenzuela 2006; Geeslin 2003; Requena and Dracos 2021, among others). The proportions of using estar as an expected response did not reach 75% for the SPAH group and they were around 50% in the L2SPA group. It has been proposed that, from an aspectual distinction, estar is the marked form as opposed to ser, which is assumed to be the default form. (Schmitt 1992; Schmitt and Miller 2007). From the point of view of bilingual acquisition, this marking condition makes forms of estar especially complex and less stable (Guijarro-Fuentes et al. 2023). The by-item analysis that was carried out, taking frequency symmetry into account, revealed a more systematic presence precisely in the estar-preferred items, which agrees with previous studies acknowledging frequency effects in HS grammars in the acquisition of aspectual morphology (Thane 2023). It could be understood that the production of the most vulnerable part of the Spanish copula pair (estar) is more permeable to the conditions presented by the input in the language, such as frequency. On the contrary, a systematic effect of frequency was not observed in the production of ser. This difference in effects for each of the members of the selection can be explained based on the proposal by Putnam and Sánchez (2013, p. 489), which foresees fluctuations in the development of HS grammars, derived from different levels of language ability activation, which would account for the wider range of variation across speakers. The restructuring of featural mapping from dominant language onto heritage language is, besides the level of competence or the frequency of the input, one of the conditions that modulate the levels of activation. This condition was concerned with the second research question, which aimed at delimiting the effect of the distinction between short and long form Russian adjectives on the production of Spanish copula choice.

In this paper, the effect of L1/dominant language is operationalized by establishing, on the one hand, interlinguistic convergence scenarios (ser + long forms; estar + short forms), which could cause a facilitative effect of transfer. On the other hand, divergence scenarios where a lower production of expected responses could be attributed precisely to the mismatch between Russian and Spanish mechanisms could also be confirmed. Concerning ser-preferred contexts, the effect of the type of interlinguistic scenario is registered in both groups of participants. For SPAH, the effect is significantly more pronounced than for the L2SPA group, due to the ceiling effect that takes place in convergence scenarios. This result could be explained by the level of competence and not necessarily by the facilitative effect of dominant language. However, the negative transfer effect seems more persistent, given that, in both groups, the items containing adjectives in the short form are associated with lower proportions of expected responses. Among the estar-preferred contexts, the results of the SPAH group reflect behavior in line with the predictions of interlinguistic influence. It can therefore be assumed that the heritage speakers in our study can restructure their L1 based upon the feature mapping of their dominant language (Putnam and Sánchez 2013). Language transfer in HS grammars has been attested at several levels of the grammar, namely phonology, morphology, syntax, semantics, and lexicon (Polinsky 2018), as well as those regarding linguistic interface phenomena (Cuza and Frank 2014; Cuza et al. 2013; Montrul 2010; Montrul and Ionin 2010; Sánchez et al. 2023). If the results by interlinguistic scenario are compared between ser-preferred and estar-preferred contexts, the effect of dominant language transfer for heritage speakers seems to be mediated by the level of mastery they show of the linguistic phenomenon in the heritage language. In this sense, our results point in the same direction as Romano (2021), when he stated that, at very advanced levels of proficiency, L2 and HL grammars are not affected by transfer from the dominant/native language. In our study, the use of the structure ser + adjective, which is a very stable combination among HS speakers, may reflect a partial transfer effect. It is in the results of the combination estar + adjective where the influence of Russian acquires greater weight.

For the results of the production of estar in the L2SPA group, the absence of differences between interlinguistic scenarios, as well as their behavior at chance—in both crosslinguistic conditions the predicted probability means are around 50%—seem to indicate persistent difficulties when carrying out this restructuring exercise. In contrast, the effect of L1 transfer is observed in the production of ser forms. Therefore, the incidence of correspondence between Russian and Spanish seems to emerge only when a sufficient level of command of the functioning of the construction in the second language has been reached. Thus, a threshold is established in the development of the L2 grammar from which the effects of the relationship between languages can influence: the fact that the Russian language formally represents, based on a morphological mechanism, the differences in meaning that Spanish expresses through a lexical-semantic mechanism in these predicative constructions.

Based on the discussion by (Guijarro-Fuentes et al. 2023), where the distinction between facilitative and negative cross-linguistic influence is discussed, a clear facilitative effect of adjectives in the long form combined with ser can be verified. This contrasts with the effect of adjectives in the short form. The low production of estar in the L2SPA group may indicate that the main difficulties for this group in the process of learning the combination estar + adjective are of a lexical nature (Guijarro-Fuentes et al. 2023). In other words, L2 learners must learn the semantic features of Spanish adjectives, which do not always coincide with Russian adjectives, based on the partial correspondence between both languages. The task of incorporating grammatical phenomena from L2 that does not completely coincide with the equivalent in L1 involves a restructuring of feature reassembly (Lardiere 2009). The effect of frequency symmetry for the ser-preferred items also in this group is consistent with the proposal that lexical knowledge is partially conditioning the domain of the Spanish copula choice. However, the effect of lexical frequency is not as stable and homogeneous as in other optionality phenomena such as modal selection in Spanish (Gudmestad 2014; Giancaspro 2020, among others).

From the general performance of the production of ser and estar between the two groups, the results confirm that the heritage speakers outperform the L2 group. This suggests a possible benefit of earlier exposure and use of Spanish. Similar advantages have been found in the domain of syntactic and syntactic-discursive phenomena (Cuza and Frank 2014; Montrul 2010). In summary, the results obtained can confirm that both cross-linguistic influence and age of onset of bilingualism are key aspects in the linguistic development of a L2 and a HL that have an impact on the use of Spanish copula choice.

7. Conclusions

The present study delimited the competence of L2 Spanish, L1 Russian speakers and Russian dominant, Spanish heritage speakers in the use of the Spanish copula choice in predicative constructions. We applied a crosslinguistic perspective that allowed us to set four different conditions and delve into the L1 transfer effects in the competence of heritage and L2 Spanish, L1 Russian speakers. According to our results, differences in the acquisitions patterns of the Spanish copula choice could be ascertained in the predicative construction. More specifically, the selection of ser appears to be more stable than the selection of estar in the grammar of the two groups of speakers. Additionally, a transfer effect constrained to speakers’ competence level was confirmed. Though promising, we believe that more research involving different tasks, linguistic structures, and features as well as different profiles of speakers belonging to different Spanish varieties is needed to be able to confirm the results delineated in this study. All in all, this study contributes to the description of the linguistic competence of an underrepresented population in heritage language research, where most research has been undertaken in English-dominant contexts.