Overt pronouns are challenging: subject realization in written narratives of Portuguese-French bilingual school-age children

RomanicaCracoviensia 2024: 383–398 https://doi.org/10.4467/20843917RC.24.029.20984 www.ejournals.eu/Romanica-Cracoviensia  https://orcid.org/0000-0001-5629-9556  https://orcid.org/0000-0001-9378-9079 CristinaFlores University of Minho [email protected] EstherRinke Goethe University Frankfurt esther[email protected] OVERT PRONOUNS ARE CHALLENGING: SUBJECT REALIZATION IN WRITTEN NARRATIVES OF PORTUGUESE-FRENCH BILINGUAL SCHOOLAGE CHILDREN ABSTRACT This paper examines the use of referential expressions in the subject position in European Portuguese (EP) in a corpus of narratives written by two groups of Portuguese-speaking school-age children: monolingually-raised children living in Portugal and bilingual French-Portuguese children living in Switzerland. We focus on the children’s choice of null and overt subjects, considering both the syntactic context (interand intrasentential) and the pragmatic context (topic continuity and topic shift). Additionally, we explore potential age and proficiency effects within the group of bilinguals. Results show an overuse of overt pronouns in intersentential topic continuity contexts in the narratives of the bilingual children, modulated by proficiency and age. This confirms that potential changes to heritage language grammars in EP may affect, in particular, the use of overt subject pronouns. The overuse of strong pronouns reflects that their acquisition is challenging for monolingual and for bilingual children alike, probably due to the complexity and variability of overt subject pronouns in the adult grammar. KEYWORDS: null subjects, overt pronouns, syntactic context, pragmatic constraints, European Portuguese, heritage languages INTRODUCTION This paper investigates the distribution of null and overt subject pronouns in narratives written by two groups of children acquiring European Portuguese (EP) as first language. The first group consists of monolingually-raised children living in Portugal, the second Cristina Flores, Esther Rinke 384 group of bilingually-raised children living in the French-speaking region of Switzerland who grow up with French as dominant environmental language and EP as heritage language (HL). We focus on the children’s use of null and overt subjects and investigate the role of syntactic (intravs. intersentential) and pragmatic (topic continuity/TC vs. topic shift/TS) context. Since the group of bilingual children is rather heterogeneous concerning age and proficiency, we also consider potential effects of these two factors on their individual results. SUBJECT REALIZATION IN EP AND FRENCH In a null subject language (NSL) like EP, the choice between different referential expressions depends on the discourse referent’s degree of accessibility, which in turn depends on factors like the syntactic function of the antecedent, the distance between antecedent and null/overt subject and the prominence of the antecedents in the previous discourse (TC/TS), among others (see Calabrese 1986; Ariel 1990; Carminati 2002). Null subjects are more likely to be used with more accessible referents, i.e., antecedents in subject position and less distant or recently mentioned antecedents, and in contexts of TC. Overt subject pronouns are more likely with less accessible referents, i.e., antecedents in object position and more distant antecedents, and in contexts of TS. Chomsky (1981) establishes a general principle guiding pronoun choice in NSLs, the “Avoid pronoun principle”, suggesting that a null pronoun is the unmarked choice, and an overt pronoun should only be used, if necessary, in cases of emphasis, contrast or ambiguity. Accordingly, corpus studies generally show a higher proportion of null than of overt subjects (typically a distribution of approx. 30% overt vs. 70% null subjects). Based on a corpus of written newspaper interviews, Barbosa et al. (2005) report that 3rd person overt and null pronoun subjects in EP are indeed realized along these lines: 22% of the subjects were realised by an overt pronoun and 78% were null subjects. Subject realisation was influenced by the position and distance of the antecedent: the lowest number of overt subjects occurred in intrasentential anaphoric contexts in which the subject of a subordinate clause referred to the subject of an adjacent main clause (3%) and in intersentential contexts in which the subject of a main clause relates to a subject antecedent in an immediately adjacent preceding main clause (11%). More overt pronouns occurred in contexts with a non-subject antecedent in an adjacent clause (29%) and in contexts with an antecedent in a non-adjacent clause (33%). This shows that overt subject use correlates with topic shift and greater distance of the antecedent. Data from sentence comprehension show a similar pattern: Rinke and Flores (2018) compare different conditions, each one either with an overt or a null pronoun (pro), see ex. (1a) for an intrasentential anaphoric and (1b) for an intersentential context. This study included adult speakers of EP and different groups of monolinguallyand bilinguallyraised child speakers. Overt Pronouns Are Challenging: Subject Realization in Written Narratives…385 (1) a. Porisso, a mãe chamoua Sónia quando therefore the mother called the Sónia when pro/ela voltou dotrabalho. pro/she came back from-the work ‘Therefore the mother called Sónia when she arrived from work.’ b. A Sónia telefonou  Susana.Como sempre pro/ela estáatrasada. the Sónia called to+the Susana As always pro/she is late ‘Sónia called Susana. As usual she is late.’ The results of the adult group indicated that in all contexts, null subjects show a strong tendency to refer to the subject in the adjacent clause (=TC, 88.9% in the intrasentential and intersentential anaphoric conditions). Overt pronouns preferentially relate to the nonsubject constituent in the adjacent clause (=TS, varying between 82.9% in the anaphoric intrasentential and 72.2% in the intersentential condition). The results reveal that in ambiguous contexts with a subject and an object antecedent, overt and null pronouns show different patterns of anaphoric resolution: null pronouns refer to a subject antecedent in an adjacent clause, overt pronouns to an object referent. This effect is widely independent of condition with respect to null pronouns. For overt pronouns, the syntactic context plays some role: although they preferentially mark topic shift, this tendency is stronger in intrasentential than in intersentential contexts (see also Morgado 2013; Lobo and Silva 2016). French is generally considered to be a non-null subject language (Kayne 1980, Rizzi 1986)1 in the sense that a pronominal subject is realized by a subject clitic, which can be doubled by strong pronouns in contexts of emphasis.2 (2) Alors la mère atéléphoné  Sonia quand therefore the mother called to-the Sonia when (*pro)/elleest rentrée du travail. pro/ she is come back from-the work ‘Therefore the mother called Sonia when she arrived from work.’ SUBJECT REALIZATION IN CHILD AND ADULT HERITAGE SPEAKERS Previous research has shown that the adequate choice between overt and null subjects in NSLs may be a vulnerable domain in HLs, both in comprehension and production (Montrul 2004; Keating et al. 2011). A general tendency observed in adult heritage speakers (HSs) is overexplicitness through the overuse of overt subject pronouns or their 1 For the assumption that French subject clitics are comparable to agreement markers in prototypical null subject languages see Kaiser (1992). 2 For a corpus study on subject realization in colloquial French, see Gabriel and Rinke (2010). Cristina Flores, Esther Rinke 386 interpretation in terms of TC (Montrul 20043; Otheguy et al. 2007). This tendency has also been reported for bilingual children with a non-dominant NSL (e.g., Sánchez et al. 2023, for Spanish-English; Serratrice et al. 2004, for Italian-English, Argyri and Sorace 2007, for Greek-English bilinguals), suggesting that changes to the constraints that regulate subject realization may be a consistent particularity of the HL, at least in some HS populations (e.g., Brehmer et al. 2023 did not attest this tendency in German-Polish bilingual children). Potential changes are often explained as the result of a combination of independent or related factors, such as cross-linguistic influence (CLI, e.g., Serratrice et al. 2004), proficiency and language dominance (e.g., Andreou et al. 2023; Argyri and Sorace 2007; Torregrossa et al. 2021). Flores and Rinke (2020) investigated the realization of overt and null subjects in a spontaneous speech corpus of two generations of EP speakers living in Germany (1st and 2 nd generation speakers with migration background). This study revealed quantitative results which were similar to the overall numbers reported by Barbosa et al. (2005): 67.1% of overt and 32.1% of null subjects in Gen1 and 66.5% of overt and 33.5% of null subjects in Gen2. Both groups also showed a sensitivity to the pragmatic context, with a clear tendency to use null subjects in TC contexts (80.6% in Gen1; 77.7% in Gen2), and a preference for overt subjects in TS contexts (64.4% in Gen1 and 55.8% in Gen2). Both groups also show the expected tendency for realizing a subject antecedent by a null subject and an object antecedent by an overt pronoun. However, the study also reports a considerable degree of individual variation in Gen2, with two speakers diverging from the general pattern and showing a tendency towards an overuse of overt subject pronouns. Concerning the comprehension of null and overt subjects, the study by Rinke and Flores (2018) showed that monolingually-raised EP, bilingual Spanish-EP and bilingual German-EP children also preferentially interpret null subjects in terms of TC and overt subjects as TS. This confirms earlier studies showing that monolingual children acquired the distribution of null and overt subjects at the age of 9 (Lobo and Silva 2016); and adds that bilingual children, independent of their language combination, also have acquired this knowledge at roughly the same age. There are, however, also some differences between the groups. In the null subject condition, the bilinguals show more variation in the interpretation of the null pronoun in the intersentential and the intrasentential cataphoric conditions, while in the overt pronoun condition monolingual and bilingual children behave alike and differ in all conditions from the monolingual adults. The latter confirms that the pragmatically adequate interpretation of overt subject pronouns is not only a challenge for bilingual children, but also a complex task in monolingual acquisition. 3 Montrul 2004 also reports a pragmatically inadequate use of null subjects in adult Spanish HSs. Overt Pronouns Are Challenging: Subject Realization in Written Narratives…387 THE PRESENT STUDY RESEARCH QUESTIONS Departing from the evidence gathered so far from studies on subject expression in NSLs acquired as HL, four research questions guide the present analysis: I. Do children who acquired EP as L1 in a monolingual or in a bilingual context show the overall distribution of null/overt subject pronouns identified for EP in their (written) narratives? II. What is the impact of the syntactic context (inter-/intrasentential) and the pragmatic constraints (TC/TS) on this distribution? III. Do bilingual children show a tendency to be overexplicit in TC contexts? IV. To what extent is the performance of bilingual children modulated by current age and proficiency? PARTICIPANTS Data were collected from two groups of Portuguese children. The monolingually raised group (MON) comprises 30 children without a migration background who reside in the Lisbon region of Portugal. They range in age from 11 to 12 years (mean age 142.5 months). These children were born into Portuguese-speaking families with exclusive exposure to EP. Their narratives were collected as part of the study by Flores et al. (2024). The bilingually raised group (BIL) consists of 32 bilingual children who live in the French-speaking part of Switzerland. They are simultaneous or early successive bilinguals who started to acquire EP as their HL from birth within the family setting and French, the societal language, either from birth, also through the nuclear family, or up to age 4 in the kindergarten (mean age of onset of French = 19.00 months, SD = 26.49). All children have Portuguese parents who were born in Portugal and migrated to Switzerland in adulthood. The bilingual children are between 10 and 12 years old, with a mean age of 133.57 months (SD = 9.12). All children attend a public French-speaking school. Additionally, they attend an extra-curricular Portuguese HL course organized by the Camões Institute once a week, lasting approximately two hours. This means that all bilingual children are mainly instructed in French, but they also acquired literacy skills in Portuguese through the HL course. As for language use, the bilingual children use EP at home, with Portuguese friends, neighbours, and other family members, and during the Portuguese HL course. They frequently visit Portugal during their holidays. METHOD Data collection in Switzerland included three types of instruments: a background questionnaire, a cloze test in EP (and in French) and a written retelling task in EP (and also in French). For the present study, we will report only on the Portuguese data (for more information about the data see Flores et al. 2022). Data collection in Portugal included Cristina Flores, Esther Rinke 388 the same retelling prompt and a short background questionnaire. Data collection took place in person in a classroom in the presence of an instructor. To elicit written data, a combination of audio and visual stimuli was used. A model story (based on an ENNI picture narrative, see Schneider et al. 2005), narrated by a native Portuguese speaker, was aurally played alongside a sequence of pictures in a PowerPoint presentation. Afterwards, children were asked to write down the story they had just heard (once), using a paper featuring the same set of images. The recorded story consisted of 258 words. The narratives produced by the monolingual children had an average of 138.66 words and those of the bilinguals had 119.12 words on average. The complete corpus encompassed 7853 words. WRITTEN NARRATIVES: CODING AND ANALYSIS The narratives were transcribed and coded for a bunch of grammatical features. For this analysis, the relevant ones are: i. subject form: definite noun phrase (dnp), indefinite noun phrase (inp), personal pronoun (pron), demonstrative pronoun (dem), relative pronoun (rel), null pronoun (null). Subject omissions in coordinated structures of the type “She entered the room and Ø switched on the light.” were coded separately; ii. pragmatic context: topic continuity (TC), topic shift (TS); iii. syntactic context: intersentential (INTER), intrasentential (INTRA).4 Example (3) shows a sentence with an overt pronoun in an intrasentential context (ele), which refers to the indirect object of the main clause (elefante), being thus a TS context. Example (4) shows a null subject (pro) in an intrasentential context which refers to the subject of the main clause (Girafito, =TC). In example (5) the overt pronoun ele refers to a subject antecedent in a previous sentence (intersentential context, TC). (3) A elefantina foi acorrerpedir ajuda ao elefantei, the elefantina went running asking help to+the elephant  para ver se eleiconseguia. to see if he was+able ‘Elefantina began to run to ask the elephant for help in order to see if he would be able.’ (4) O Girafito ficoumuitofuriosoapensarque proiriaperderoseubrinquedo. the Girafito was very angry to think that would lose the his toy ‘Girafito became very angry just thinking that he would lose his toy.’ (5) (E o girafinho ficou muito triste.) Ele ficou tão triste que... he became so sad that ‘And Girafinho became very sad. He became so sad that….’ 4 We further coded for intrasentential anaphoric and cataphoric contexts, however, due to the limited occurrences of cataphoric contexts, we collapsed both. Overt Pronouns Are Challenging: Subject Realization in Written Narratives…389 BACKGROUND QUESTIONNAIRE AND PROFICIENCY TEST In addition to the written narratives, a background questionnaire was filled out by the parents of the bilingual children to gather information on language acquisition and language use in various contexts of the children’s life. Furthermore, data collection comprised a written cloze test in EP, which served as proficiency assessment tool. The test included 40 blank spaces targeting various linguistic structures of different levels of difficulty (for more details see Torregrossa et al. 2023). RESULTS DESCRIPTIVE RESULTS We will start by reporting the descriptive analysis of the corpus. Table 1 shows the raw counts of all subject forms used by the monolingual and the bilingual children in their written narratives. The total number of coded subjects per group is similar in both groups (MON: 645; BIL: 640). Table 1. Types of subject expression (raw counts) Group Definite NP Overt pronoun Null pronoun Indefinite NP Relative pronoun Demonstrative Coord. null Total Monolingual 276 78 114 58 37 2 80 645 Bilingual 305 119 62 61 23 0 70 640 In a second step, we circumscribed our analysis only to the pronominal subject forms (see Table 2). Table 2. Types of subject expression in 3rd person contexts (raw counts) Group Subject form Raw counts (%) Monolingual Null 114 (59.4%) Pron 78 (40.6%) Bilingual Null 62 (34.3%) Pron 119 (65.7%) As seen in Table 2, the proportion of overt subjects is higher in the bilingual group and the proportion of null subjects lower. While in MON the rate of null subjects (59.4%) surpasses the rate of overt pronouns (40.6%), in BIL the proportion of overt pronouns (65.7%) exceeds that of null subjects (34.3%). Cristina Flores, Esther Rinke 390 Table 3 gives an overview over the distribution of null and overt pronouns per syntactic and pragmatic context. Table 3. Null and overt subjects per syntactic and pragmatic context (raw counts and %) Group Subject form Raw counts (%) Monolingual Intersentential TC null 51 (62.2%) pron 31 (37.8%) TS null 15 (30.6%) pron 34 (69.4%) Intrasentential TC null 28 (80%) pron 7 (20%) TS null 4 (80%) pron 1 (20%) Bilingual Intersentential TC null 27 (25.5%) pron 79 (74.5%) TS null 12 (28.6%) pron 30 (71.4%) Intrasentential TC Null 9 (69.2%) pron 4 (30.8%) TS Null 0 (0%) pron 2 (100%) STATISTICAL ANALYSIS Statistical analyses were performed on the jamovi script (The jamovi project 2021) in R (R Core Team 2021). Since our focus lies on the alternation between null and overt pronouns, for the statistical analysis we excluded the definite noun phrases and reduced the variable ‘subject type’ to a binominal dependent variable with two factors, ‘null’ and ‘pronoun’ (pronoun is the reference level). We ran a generalized linear mixed model (GLMM) with the following independent variables: ‘group’ (MON/BIL), ‘pragmatic context’ (TC/TS) and ‘syntactic context’ (INTER/INTRA), which were simple coded (–0.5/+0.5). The model included also simple interactions of the independent linguistic variables with the variable ‘group’ and participants as random intercepts. The test results are displayed in Table 4. Overt Pronouns Are Challenging: Subject Realization in Written Narratives…391 Table 4. GLMM results Names Estimate SE exp(B) Lower Upper ZP (Intercept) –0.0286 0.225 0.972 0.625 1.510 –0.127 0.899 Group –0.8597 0.459 0.423 0.172 1.041 –1.871 0.061 TC/TS 0.7697 0.292 2.159 1.217 3.830 2.632 0.008 Syntactic_context –1.2378 0.379 0.290 0.138 0.609 –3.270 0 .001 Group*TC/TS 1.3357 0.581 3.803 1.217 11.878 2.298 0.022 Group* Syntactic_context 0.3743 0.756 1.454 0.331 6.394 0.495 0.620 The GLMM shows a significant effect of ‘group’, of the ‘pragmatic context’ (TC/TS) and of the ‘syntactic context’ in the distribution of null/overt forms in the corpus. The negative estimate for ‘group’ indicates that the log odds of producing overt pronouns decreases for monolingual children, while for ‘syntactic context’ it shows that the log odds of producing overt pronouns decreases in intrasentential contexts. The positive estimate for TC/TS indicates a tendency to produce more overt subject pronouns in TS than in TC contexts. Furthermore, there is a significant effect of the interaction between ‘group’ and ‘pragmatic context’ (TC/TS), but no interaction between ‘group’ and the ‘syntactic context’. These results indicate that, overall, the choice between an overt/null pronominal form is influenced by the pragmatic as well as by the syntactic context, although this result must be taken with caution because the bilingual children produced a low number of intrasentential contexts. In addition, results show that the two child groups differ in their choice of pronominal forms (null vs. overt), but this difference is influenced by the pragmatic context (TC vs. TS), and not by the syntactic context. To better understand the Group*TC/TS interaction, we ran Post Hoc Comparisons (with Bonferroni correction) (see Table 5). Table 5. Post Hoc Comparisons – Group*TC/TS Group:TC/TS Group:TC/TS exp(B) SE Z P BIL TC – BIL TS 0.903 0.3903 –0.236 1.000 MON TC – MON TS 0.237 0.0930 –3.672 0.001 BIL TC – MON TC 4.607 2.1500 3.273 0.006 BIL TS – MON TS 1.211 0.7400 0.314 1.000 Results of the Post Hoc Comparison indicate that monolingual and bilingual children differ significantly in their choice of pronominal forms in TC contexts (p = 0.006), but not in TS contexts. Furthermore, intragroup comparisons show that monolingual children have different choices in TC vs. TS contexts (p = 0.001), but bilingual children do not differ significantly concerning the distribution of overt vs. null forms in topic shift and topic continuity contexts. Cristina Flores, Esther Rinke 398 Torregrossa Jacopo, Flores Cristina, Rinke Esther, 2023, “What modulates the acquisition of difficult structures in a heritage language? A study on Portuguese in contact with French, German and Italian”,Bilingualism:LanguageandCognition, 26, 1, 179–192. Torregrossa Jacopo, Andreou Maria, Bongartz Christiane, Tsimpli Ianthi M., 2021, “Bilingual acquisition of reference: The role of language experience, executive functions and cross-linguistic effects”, Bilingualism:LanguageandCognition, 24, 4, 694–706.