Studies in Italian as a Heritage Language 9783110759587, 9783110759518

Citation preview

Studies in Italian as a Heritage Language

Language Contact and Bilingualism

Editor Yaron Matras

Volume 25

Studies in Italian as a Heritage Language Edited by Francesco Bryan Romano

ISBN 978-3-11-075951-8 e-ISBN (PDF) 978-3-11-075958-7 e-ISBN (EPUB) 978-3-11-075962-4 ISSN 2190-698X Library of Congress Control Number: 2022951404 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the internet at http://dnb.dnb.de. © 2023 Walter de Gruyter GmbH, Berlin/Boston Cover image: Anette Linnea Rasmus/Fotolia Typesetting: Integra Software Services Pvt. Ltd. Printing and binding: CPI books GmbH, Leck www.degruyter.com

Alla piccola Valentina ♥ In memory of the loved ones lost after the outbreak of COVID19 and during the compilation of this volume.

Acknowledgments There is a long list of individuals I need to thank for their material help and psychological support in making this book possible. I would especially like to thank the series editor, Yaron Matras, for the generous time allowance spent replying to my email, zooming me, and chaperoning me through the process of editing my first edited book. My greatest gratitude also goes out to my contributors for their constancy, dedication, perseverance, patience, and indefatigable work on their chapters and the volume. Finally, I am also especially indebted to the following individuals for their material support, in no particular order: Silvia Perpiñan, Rakel Österberg, Ada Valentini, Ad Backus, Giuliana Giusti, Pedro Guijarro-Fuentes, and Madeleine Bäck Romano.

https://doi.org/10.1515/9783110759587-202

Contents Acknowledgments  List of figures 

 XI

List of tables 

 XIII

 VII

Francesco Romano Introduction   1 Silvina Montrul Chapter 1  Heritage language development: Dominant language transfer and the sociopolitical context   5

Part I: Experimental studies Pedro Guijarro Fuentes, Iria Bello Viruega, Estela García Alcaraz, Sergio Viveros Guzmán Chapter 2  Ultimate attainment in long-immersed heritage Italian immigrants: Syntactic and semantic knowledge of direct object clitics and partitive ne   35 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa Chapter 3  Grammatical competence in adult heritage speakers of Italian and adult immigrants: A comparative study   65 Francesco Romano Chapter 4  Ultimate attainment of gender in heritage and L2 Italian  Maria Teresa Bonfatti-Sabbioni Chapter 5  Auxiliary selection in heritage speakers of Italian 

 127

 95

X 

 Contents

Jacopo Torregrossa, Irene Caloi, Andrea Listanti Chapter 6  The acquisition of syntactic structures in heritage Italian: Assessing the role of language exposure at critical periods   155 Katrin Schmitz, Tim Diaubalick Chapter 7  The expression of (deontic and epistemic) modality in Italian as heritage language in Germany   195

Part II: Observational studies Caterina Ferrini Chapter 8  Italian as heritage language in third generations on social networks: Morphosyntactic code-switching features   233 Elisa De Cristofaro, Linda Badan Chapter 9 Discourse markers in heritage Italian spoken in Flanders 

 255

Margherita Di Salvo, Eugenio Goria Chapter 10  Auxiliary selection in Italo-Romance heritage languages: Argentina and the UK   289 Francesco Romano, Antonella Sorace Conclusion and future directions  Index 

 325

 321

List of figures Chapter 1 Figure 1

External factors affecting specific linguistic features in heritage language competence and use   21

Chapter 3 Figure 1 Figure 2

Density plot of the scores on sentence repetition by group   83 Bar plot representing percentage of correct answers on production of DO clitic by group   85

Chapter 4 Figure 1 Figure 2

Figure 3

Figure 4

Gender declension classes in Italian   104 Sample trial in the priming task. Photos are taken with permission from the International Picture Naming Project (Szekely et al., 2004), Heaton (1966), and Van Patten, Lee, and Ballman (1992)   111 Accuracy in the GJT by gender and group. Only accuracy for response times > 1000 and < 5000 ms were considered. L1, monolinguals. HL, heritage speakers. L2, L2 learners. Cor = correct responses, Inc = incorrect responses   118 Response times in the GJT by gender and group. Only response times to correct responses which were cube-root transformed for normality were included   119

Chapter 5 Figure 1

Unaccusativity gradience with target verbs 

 146

Chapter 6 Figure 1

Figure 2

Figure S1

Boxplot of the distribution of the scores related to children’s exposure to Italian from different persons in and outside of the home across different age stages (0–3, 3–6 and 6 years). The scores are expressed as a percentage of the number of persons from which the child received input in Italian over the total number of persons from which the child received input in one or the other language (see Section 4.1). The values range from 0 (no exposure to Italian) to 1 (exclusive exposure to Italian)   170 Means and standard errors (± 1.5) for children’s response accuracy in association with the different types of clauses (complement, adverbial, clauses involving movement or movement and embedding) at the time of testing   172 Two consecutive pictures from the Sentence Repetition Task. The children first heard the pre-recorded sentence and then saw the corresponding picture   189

https://doi.org/10.1515/9783110759587-204

XII 

 List of figures

Chapter 9 Graph 1 Graph 2 Graph 3 Graph 4 Graph 5

comparison of the syntactic position of the DMs in the L1, L2 and HL corpora   269 comparison of the grammatical category of the DMs in the MonoL1, L2 and HL corpora   269 interaction between pragmatic macro-function and syntactic position in the MonoL1  270 corpus  interaction between pragmatic macro-function and syntactic position in the L2 corpus   272 interaction between pragmatic macro-function and syntactic position in the HL corpus   272

Chapter 10 Figure 1 Figure 2 Figure 3

Groups considered for the two case studies   301 Distribution of HAVE and BE in Heritage Piedmontese according to transitivity and generations. Percentage values   314 Distribution of HAVE and BE in Heritage Campanian according to transitivity and generations. Percentage values   314

List of tables Chapter 2 Table 1 Table 2 Table 3 Table 4 Table 5 Table 6 Table 7 Table 8 Table 9 Table 10 Table 11

Summary of property asymmetries for object drop in Italian and Castilian Spanish   42  45 Participants’ pool  Conditions tested   50 Interpretable features: Spanish monolingual speakers   52 Uninterpretable features: Spanish monolingual speakers   53 Interpretable features: Italian monolingual speakers   53 Uninterpretable features: Italian monolingual speakers   54 Interpretable features: Heritage vs. monolingual speakers of Italian   55 Uninterpretable features: Heritage vs. monolingual speakers of Italian   56 Interpretable features: Heritage speakers of Italian vs. Spanish monolingual speakers   57 Uninterpretable features: Heritage speakers of Italian vs. Spanish monolingual speakers   58

Chapter 3 Table 1

Table 2 Table 3 Table 4 Table 5

Descriptive data for the two groups, Adult Immigrants (AI) and Heritage speakers (HS): mean age in years (and SD), Age of first Exposure to Italian (AoE), level of education, mean years of formal education in Italian (and SD), mean years in the UK (and SD)   77 Examples for each type of sentence in the Comprendo task (Cecchetto et al. 2012)   78 Raw scores (and SDs) for each group in the background tasks Comprendo (Italian) and TROG-2 (English)   79 Examples of the sentences contained in the sentence repetition task, the description of the sentence in terms of length and complexity, and the predicted error   80 Mean raw scores of responses (and SDs) for each group in the experimental tasks of sentence repetition and clitic production   82

Chapter 4 Table 1 Table 2 Table 3 Table 4 Table 5 Table 6 Table 7

Gender declension classes in Swedish   107 Participant information   110 Accuracy in use of gender in the priming task   114 Pairwise comparisons for accuracy in the priming task  Breakdown of gender-related errors in the priming task  Accuracy in use of gender in the GJT   117 Pairwise comparisons for accuracy in the GJT   117

https://doi.org/10.1515/9783110759587-205

 115  116

XIV 

 List of tables

Chapter 5 Table 1 Table 2 Table 3 Table 4 Table 5 Table 6 Table 7 Table 8

Percentages of the subjects’ auxiliary selection and presence/absence of acquisition   139 Percentage of the subjects’ auxiliary selection in three semantic groups in each family nucleus and deviations   139 Presence/absence of acquisition in children and presence/absence of deviations in parents, in each family nucleus in three semantic groups   140 Subjects’ auxiliary selection and omission with target verbs   142 Subjects’ auxiliary preference   144 Subjects’ auxiliary selection with Correre (to run)   144 Overall subjects’ auxiliary preference and omission in all target verbs (24)   145 Linguistic profile of the children as heritage speakers of Italian   149

Chapter 6 Table 1 Table 2 Table 3

Table 4

Table 5

Table 6

Table S1

Table from the parental questionnaire related to the question “which languages did your child hear and speak when s/he attended kindergarten (between 3 and 6)?”   165 List of all target structures considered in the analysis as sorted out based on the type of structure (Please see Table S1 for translations)   171 Hierarchical regression model with repetition accuracy of complement clauses as dependent variable and variables related to language exposure between 0 and 3, 3 and 6 and at 6 years of age as predictors   174 Hierarchical regression model with repetition accuracy of adverbial clauses as dependent variable and variables related to language exposure between 0 and 3, 3 and 6 and at 6 years of age as predictors   175 Hierarchical regression model with repetition accuracy of structures involving movement and embedding as dependent variable and variables related to language exposure between 0 and 3, 3 and 6 and at 6 years of age as predictors   175 Hierarchical regression model with repetition accuracy of clauses involving movement and embedding as dependent variable and variables related to language exposure between 0 and 3, 3 and 6 and at 6 years of age as predictors   176 List of all sentences and target structures   187

Chapter 7 Table 1 Table 2 Table 3 Table 4 Table 5 Table 6 Table 7

Core linguistic means for different modality types in Italian and German (sorted by type of means: M = mode, T = tense, L = lexical)   213 Participant Description   216 Speech rate of monolinguals and bilingual adult speakers (words per minute, following the method by Viswanath 2013)   217 Numbers of tokens with a modal verb + infinitive-construction per group   218 Numbers of tokens with a verb form in Subjunctive Imperfect per group   221 Numbers of tokens of epistemic adverbs per group   223 Statistical results comparing the epistemic adverbs in the two groups   223

List of tables 

 XV

Chapter 9 Table 1 Table 2 Table 3 Table 4 Table 5 Table 6 Table 7

basic information and linguistic background of the HLS   262 basic information and linguistic background of the L2 speakers of Italian   263 Total amount of DMs and respective normalized frequency in the MonoL1, L2 and HL corpus   266  266 ten most frequent DMs in the MonoL1 corpus  ten most frequent DMs in the L2 corpus   267 ten most frequent DMs in the HL corpus   267 Taxonomy of discursive functions used for the tagging of discourse markers in the Nvivo software for the qualitative analysis   283

Chapter 10 Table 1 Table 2 Table 3 Table 4 Table 5 Table 6 Table 7 Table 8

Patterns of AS in Central and Southern dialects of Italy   296 Informants used for this study in Argentina and Bedford, UK   303 Distribution of HAVE and BE as past-tense auxiliaries in the Argentinian corpus, across different verb types and across two generations (N=318)   305 Distribution of HAVE and BE in the Argentinian corpus, according to Sorace’s Auxiliary Selection Hierarchy   306 Distribution of HAVE and BE in unaccusatives in the Argentinian corpus, according to person   307 Distribution of HAVE and BE as past-tense auxiliaries in the Bedford corpus, according to transitivity and in the two generations. (N=274)   309 Distribution of HAVE and BE in the Bedford corpus, according to Sorace’s Auxiliary Selection Hierarchy   309 Distribution of HAVE and BE in the Bedford corpus, according to verb type and person (percentages, N=274)   311

Francesco Romano

Introduction The Godfather trilogy, Goodfellas, The Green Book, Rocky, My cousin Vinny, or even Joey in the TV series Friends, what these titles all have in common might be easy to spot: the Italian-American culture. Such TV titles have gained Italian language and culture enormous attention and success worldwide, depicting individuals of Italian-American descent as some of the most charismatic, funny and dramatic in the history of cinema. These individuals, like many others growing up in places such as South American and Europe speaking Italian are the very subjects of the research presented in this book. In the field of linguistics, they have come to be known as a special population of bilinguals known as heritage language speakers whose linguistic abilities are described from different angles in this volume. Although research on heritage languages is by now part and parcel of mainstream SLA and bilingualism publication outlets (e.g. the special issue in Bilingualism, Language and Cognition in 2020, or the critical commentaries to Domínguez, Hicks and Slabakova, 2019, in Studies in Second Language Acquisition), considerable curiosity and interest for linguistically-informed research in heritage languages continues to proliferate. Within the general framework of heritage language research, there are very good reasons for a volume on Italian. While books on L1 and L2 acquisition of Italian are already published (Belletti and Guasti, 2015; Giacalone-Ramat, 2006), no accounts exist of research on Italian as a heritage language. Another important reason is that considerable research efforts have been paid to detailing heritage language acquisition for influential languages such as Spanish which is linguistically intimately connected to its Romance cousin, Italian. The work in this volume, therefore, represents a step forward in better understanding heritage language bilingualism by comparison to the mainstream findings of Spanish and other influential heritage languages such as Russian. In addition, as a richly inflected language, Italian presents a robust case for detailing the acquisition of inflectional morphology, a domain which has time and again shown to constitute an area of great controversy not only in heritage language acquisition but also in the area of normal, atypical, and non-native language development (see Chapters 2, 3, 4, and 8). The book brings together 9 empirical studies of Italian as a heritage language from around the world plus a state-of-the-art on heritage language research by Silvina Montrul, world-renowned scholar in the field of heritage language and linguistics, in Chapter 1. The volume bears testimony of the acquisition of Italian as a lesser-studied heritage language in the Americas (US and Argentina) and Europe (Spain, Sweden, Germany, the UK, and Belgium). Moreover, it comprises studies focused on linguistic, psycholinguistic and sociolinguistic dimensions of https://doi.org/10.1515/9783110759587-001

2 

 Francesco Romano

heritage language bilingualism which appeal to an array of linguists interested in syntax and its interfaces with morphology, semantics, and pragmatics, as well as lexico-phonology. In addition, the book strikes a balance between formal and non-formal approaches to language acquisition, work by expert as well as novel researchers, and male and female scholars. In chapter 1, Montrul surveys a wide-array of research on heritage languages, narrowing in on linguistics-informed studies and highlighting two key areas in need of further research, L1 transfer and the socio-political context. Next, a series of experimental studies are presented in Part I which opens with a chapter by Guijarro-Fuentes, Bello-Viruega, García-Alcaraz, and Viveros-Guzmán. Here the ultimate attainment of Spanish direct object drop and L1 attrition of Italian partitive ne is detailed in first generation heritage language speakers of Italian resident in Spain. The aim of the study is to test whether, unlike interpretable semantic features which have been proposed to be subject to instability (Tsimpli, 2014 and references therein), uninterpretable syntactic features are less susceptible to attrition, and easier to acquire in the L2 (e.g., Tsimpli et al, 2004; Tsimpli & Dimitrakopoulou, 2007 among others). Chapter 3 by Smith, Spelorzi, Sorace, and Garraffa explores whether linguistic markers often employed to diagnose atypical monolingual profiles in Italian children can also detect differences in the grammatical patterns of adult heritage speakers of Italian living in the UK. To follow, in chapter 4, Romano investigates the ultimate attainment of Italian gender comparing heritage and L2 speakers of Italian dominant in Swedish with native Italian monolinguals. The purpose of the study is to test two well-known hypotheses of morphological variability in the L2 literature, the Failed Functional Feature Hypothesis (Hawkins and Chan, 1997) and the Missing Surface Inflection Hypothesis (Prévost and White, 2000), extending them to heritage populations. Chapter 5 by Bonfatti-Sabbioni, continues the streak of experimental studies, exploiting a forced-choice judgment task to compare the selection of auxiliary avere and essere in heritage children and their parents. The purpose of the study is to identify similarities and differences between the grammars represented in the input that heritage speakers receive (i.e. the parents’) and the grammars eventually developed (i.e. the heritage children’s). Chapter 6 by Torregrossa, Caloi, and Listanti, expands the ever-growing and popular use of sentence repetition tasks as tools for measuring language proficiency. More specifically, their study seeks to describe the acquisition of a series of syntactic structures in Italian heritage children dominant in German with particular reference to language exposure across different age ranges. In part I’s closing chapter 7, Schmitz and Diaubalick hone in on a fairly understudied linguistic phenomenon, that of the expression of modality and the semantic-pragmatics interface. Semi-structured interviews are employed with monolingual Italians and bilingual German-Italian adults to discuss similarities and differences in the way

Introduction 

 3

these two populations realise modality values paying particular attention to the role of transfer from the dominant to the heritage language. Part II presents three observational studies focusing on qualitative analyses. In chapter 8, Ferrini describes the spoken-typed language of heritage speakers as a form of digital communication. To be more specific, she presents a qualitative analysis of the language used by third generation heritage language speakers in the USA on three Facebook groups. The analysis examines 2150 cases of linguistic contact from a code-switching interactional perspective. In turn, chapter 9 by De Cristofaro and Badan, like Romano’s chapter 4, compares the performance of heritage speakers with L2 speakers and native Italian monolinguals, focusing this time on both quantitative and qualitative aspects of the use of discourse markers in Italian. Set in Belgium, the study details the elicitation of typology, frequency, pragmatic functions, and syntactic positions of discourse markers, perspectivised by code-switching. In the final chapter of the book, chapter 10, Di Salvo and Goria, much like Bonfatti-Sabbioni in chapter 5, narrow in on auxiliary selection (AS). However, the authors extend their interest to two different Italo-Romance varieties used as heritage languages in two migratory settings: the Piedmontese dialect in the regions of Córdoba and Santa Fé (Argentina) and the Campanian dialect of Montefalcione (Campania, AV, South of Italy) in the English city of Bedford. This design enables them to investigate the effects of language contact between systems with split intransitivity, namely the Italo-Romance varieties of Piedmontese and Montefalcione, and without, chiefly the Spanish of Argentina and English of the UK, evaluating similarities and differences between the two scenarios.  A few important disclaimers. First, chapter 10 is a chapter some may disagree to be purely representative of Italian as a heritage language rather than ItaloRomance dialects. Although Italian dialects are, by UNESCO standards, languages in their own right in similar ways to other official languages spoken in de facto or nearly-sovereign territories within a larger nation (e.g. Catalan and Galitian in Spain), the dialects presented in chapter 10 are linked to the notion of “Italianness” both in linguistic and cultural heritage as the authors describe in the chapter. The chapter was additionally included due to the sizeable interest in Italian as a heritage language. While studies of Spanish as a heritage language, for example, have proliferated in the field of heritage bilingualism, far less research has been spent on Italian, mostly due to the reduced popularity and presence of the language compared to its Romance cousin. Next is chapter 2 where the authors investigate attrition in first-generation Italian immigrants. A few words are in order here as disagreement exists in terms of qualifying “heritage”. As editor to this volume, I do not wish to favor any one definition of heritage speaker over another (Montrul, 2016; Polinsky, 2021; Kupisch and Rothman, 2018, inter alia) but instead would like the volume to promote fervent discussion and constitute a forum for productive

4 

 Francesco Romano

disagreement in order to advance the field of heritage language bilingualism. This volume, thus, is ultimately a space for less common ideas to come to the fore and challenge the status-quo.

References Belletti, Adriana & Maria Teresa Guasti. 2015. The Acquisition of Italian: Morphosyntax and Its Interfaces in Different Modes of Acquisition. Amsterdam: John Benjamins. Dominguez, Laura, Glyn Hicks & Roumyana Slabakova. 2019. Terminology choice in generative acquisition research: the case of “incomplete acquisition” in heritage language grammars. Studies in Second Language Acquisition 41(2). 241–255. Giacalone Ramat, Anna (ed.). 2006. Verso l’italiano: percorsi e strategie di acquisizione. Roma: Carocci. Hawkins, Roger & Cecilia Yuet-hung Chan. 1997. The partial availability of Universal Grammar in second language acquisition: The ‘failed functional features hypothesis’. Second Language Research 13. 187–226. Kupisch, Tania & Jason Rothman. 2018. Terminology matters! Why difference is not incompleteness and how early child bilinguals are heritage speakers. International Journal of Bilingualism 22(5). 564–582. https://doi.org/10.1177/1367006916654355 Montrul, Silvina. 2016. The Acquisition of Heritage Languages. Cambridge University Press. Polinsky, Maria. 2018. Heritage Languages and Their Speakers. Cambridge University Press. Polinsky, Maria & Gregory Scontras. 2020. Understanding heritage languages. Bilingualism: Language and Cognition 23(1). 4–20. doi:10.1017/S1366728919000245 Prévost, Philippe & Lidia White, L. 2000. Missing surface inflection or impairment in second language acquisition? Evidence from tense and agreement. Second Language Research 16. 103–33. Tsimpli, Ianthi Maria. 2014. Early, late or very late? Timing acquisition and bilingualism. Linguistic Approaches to Bilingualism 4. 283–313. Tsimpli, Ianthi, Antonella Sorace, Caroline Heycock & Francesca Filiaci. (2004). First language attrition and syntactic subjects: a study of Greek and Italian near-native speakers of English. International Journal of Bilingualism 8. 257–277. Tsimpli, Ianthi Maria & Maria Dimitrakopoulou. 2007. The Interpretability Hypothesis: Evidence from wh-interrogatives in second language acquisition. Second Language Research 23. 215–242.

Silvina Montrul

Chapter 1  Heritage language development: Dominant language transfer and the sociopolitical context Abstract: Heritage languages are sociopolitically minority languages learned in a bilingual or multilingual a context. They include standard national languages in the diaspora spoken by immigrants and their children, among other situations. Heritage speakers are bilinguals (or multilinguals) who tend to be dominant in the majority language and their proficiency in the heritage language varies from merely receptive to fully fluent. Place of birth, age at immigration and age of onset of bilingualism are key variables that determine language proficiency in the heritage language. Linguistics-based experimental studies of heritage languages have focused on understanding both current and developing linguistic knowledge of heritage speakers in different grammatical areas. Collective findings from studies on different heritage languages allow generalizations about what has been uncovered so far and identification of gaps in our current knowledge. This chapter provides a general overview of heritage languages within linguistics research and discusses areas in need of further research, such as the role of dominant language transfer and sociopolitical factors that contribute to eventual proficiency, language maintenance or loss, in heritage languages. Keywords: heritage language, second language, majority language transfer, phonology, morphosyntax, Italian, Spanish, sociopolitical context

1 Heritage languages and heritage speakers Heritage languages have come to be defined as socio-politically minority languages that develop in language contact situations and are acquired since birth in a bilingual or multilingual setting (Montrul and Polinsky 2021). Heritage languages include indigenous languages in many countries and territories (e.g., Dyirbal in Australia, Guaraní in Paraguay, Quechua in Peru, Bolivia and Ecuador, Sammi in Norway and Sweden), historical minority languages with official or non-official status in their territories (e.g., Catalan in Catalunya, Euskera in the Basque Country, Irish in Ireland, Welsh in Wales) when acquired outside their territories, as well as languages with https://doi.org/10.1515/9783110759587-002

6 

 Silvina Montrul

less or no official status within a nation (e.g., Euskera and Breton in France), and immigrant languages brought by the thousands of migrants that have arrived and continue to arrive in different countries. Although many languages have a homeland (e.g. Italian, English, French, Korean, Spanish, Russian, Hebrew), are learned in a monolingual environment by children, and have official status within their territories, these same languages can be a heritage language in a diasporic situation. For example, Spanish is the main official language of Spain and nineteen countries in Latin America, but it is also a heritage language in the United States and in many countries of Europe. Similarly, although English is not often thought of as a heritage language because of its international status and prestige, it is learned as a heritage language in Israel, in Japan, and in Argentina, to give just a few examples. Italian, the official language of Italy, is spoken as a heritage language in Canada, the United States, the United Kingdom, Brazil, Argentina, Germany, Sweden, Flanders, as represented in many of the chapters in this volume (Romano 2022). Heritage speakers are bilingual or multilingual speakers who were exposed to the heritage language at home from birth exclusively or together with the majority language in early childhood, but by adulthood they tend to be dominant in the majority language (Lloyd-Smith et al. 2020). Many contextual variables contribute to the range of proficiency (from minimal to fully fluent) observed in the heritage language: Place of birth (homeland vs. diaspora), age at immigration and age of onset of bilingualism, size of the heritage language community, availability and vitality of the heritage language beyond the home, including schooling, among many other factors, contribute to the degree of ultimate attainment in the heritage language. A distinction is often made about generation: the immigrants who grew up in a monolingual environment and emigrated in adulthood are considered first-generation, their children second-generation, and their grandchildren third-generation. The first-generation speakers usually keep and maintain their native language as the dominant language and pass it onto their children, who grow up in a bilingual environment. The second-generation speakers grow up in a bilingual situation. Regardless of whether the heritage language is acquired since birth, simultaneously with the majority language (as in children of immigrants born in the host country), or as a first and only language before immigration (as in immigrant children who arrive with their parents and learn the majority language as a second language), heritage speakers show different degrees of command of the heritage language in adulthood. By the third and fourth generations, the majority language is typically dominant. The active maintenance of the heritage language becomes more unlikely in the third and fourth generations, although this depends on the language, the sociopolitical context, and the community who speak it. Most studies published to date tend to focus on first generation and secondgeneration immigrants. The linguistic knowledge and potential language change

Chapter 1 Heritage language development 

 7

of first-generation immigrants, who immigrated in adulthood, is the main focus of the field of native language attrition. Second generation immigrants are the most studied cases of heritage speakers, because this group displays a wide range of variation in their knowledge and use of the heritage language. Several chapters in the present volume are about first-generation Italian immigrants (Guijarro Fuentes et al. 2022), second generation heritage speakers of Italian (Bonfatti-Sabbioni 2022, Romano 2022, Torregrossa and Listani 2022, Schmitz and Diaubalick 2022), and a comparison between the two groups (Smith et al. 2022). Smith et al. (2022) show that the heritage speakers performed significantly worse than first generation immigrants in a sentence repetition task targeting several morphosyntactic structures of Italian, a finding consistent with previous observations. The study of heritage languages is not new. Heritage languages have been a prominent topic in sociolinguistics, anthropological linguistics and folklore studies. What has changed in the last twenty-five years, however, is that the theoretical and experimental study of heritage languages have made it to the forefront of linguistic research. With close to 6,900 living languages worldwide distributed in 200 countries, most of the world population is bilingual or multilingual (i.e., they have knowledge of and use two or more languages to different degrees) and there is growing recognition that our theories of language are no longer inclusive and representative of our reality because they are based on idealized models of monolingual native speakers, and mostly on English. Because heritage languages develop under conditions of reduced input and use during childhood, they are a critical test case for understanding the relationship between language acquisition and human language variation; more specifically, how the role of input in language development affects language representation and processing in adulthood. With respect to the modularity of linguistic knowledge, the study of heritage languages reveals how selected aspects of the heritage grammar may or may not be vulnerable to majority language transfer, simplification, and eventual language change. Experimental studies of heritage languages have focused on understanding both current and developing linguistic knowledge of heritage speakers in different grammatical areas. The collective findings from many studies on different heritage languages allow us to make some generalizations about what we have uncovered so far and to identify gaps in our current knowledge for further research. In this chapter I first give a general overview of heritage languages and what linguistic research has revealed, including some of the new studies presented in this volume. In the second part, I discuss areas in need of further research, such as the role of dominant language transfer and sociopolitical factors that contribute to eventual proficiency, language maintenance or loss, in heritage languages. Although these two factors can be confounded, I discuss ways in which they can be investigated in future work.

8 

 Silvina Montrul

2 The grammars of heritage speakers Even though heritage speakers vary on many dimensions (language, territory, age of acquisition, language and languages in contact, proficiency, availability of the language beyond the home, schooling in the heritage language, among many others) their grammars show common patterns in different areas of linguistic knowledge, such as morphology, syntax, semantics and phonology. Some aspects of heritage speakers’ grammars converge with the grammars of fluent speakers of the language, who did not grow up in a language contact situation, while many others diverge in specific ways. Accumulated evidence has shown that, compared to second language learners, heritage speakers have superior phonological abilities in speech perception and speech production (Chang et al. 2011, Gor 2014, Kim 2020, Lukyanchenko and Gor 2011). Studies of speech perception have found that heritage speakers do not differ from baseline speakers, whereas in production, however, heritage speakers may have perceptible accents. Lukyanchenko and Gor (2011) tested phonological perception of soft consonants in Russian (/t-t’/ and /p-p’/ contrasts) and found that the heritage speakers were more accurate than L2 learners of Russian and did not differ from Russian native speakers. Chang (2016) examined Korean heritage speakers’ phonological perception of syllable final stops in Korean and in English and found that the heritage speakers were indistinguishable from the Korean native speakers: they even surpassed the perceptual abilities of English native speakers in English. Kim (2020) tested perception and production of lexical stress in Spanish, which is contrastive with some verb forms (canto ‘I sing’, cantó, ‘she/he sang’), in nuclear position, prenuclear position and in unaccented contexts. In perception, the heritage speakers did not differ statistically from the native speakers; in production however, the heritage speakers patterned with the L2 learners in producing non-target stress instead of with the native speakers. Kupisch et al. (2014) and Lloyd-Smith et al. (2020) have also shown that in production, heritage speakers of French and Italian do not always pattern with baseline speakers, displaying a “heritage accent”. In sum, heritage speakers show native phonological perceptual abilities in the heritage language and these abilities are likely due to very early exposure to the heritage language in infancy. There are age effects in the perception and production of phonology: the younger the exposure to the majority language (the L2) the less native the ability in the heritage language (Ahn et al. 2017, Amengual 2019, Montrul 2008). Native production not only requires early onset of acquisition but also language use (Knightly et al., 2003). In the area of syntax, Håkansson’s (1995) study of Swedish heritage speakers found that despite significant problems with gender morphology, these participants had solid knowledge of V2 in the clausal domain. Other studies showing retention

Chapter 1 Heritage language development 

 9

of V2 in heritage Scandinavian are Larsson and Johannessen (2015) and Eide and Hjelde (2015), although Polinsky (2018) discusses cases where V2 is less preserved in structures involving the complementizer phrase (CP structures). Studies of Spanish and Italian heritage speakers have found that the syntax of clitic pronouns, especially the position of clitics with respect to the finiteness of the verb, is well acquired and maintained, and there is evidence of convergence with native norms (Montrul 2010, Romano 2020). As reported by Polinsky (2018), studies of Spanish, Korean and Japanese heritage speakers found robust knowledge of unaccusativity (Fukuda 2017, Lee 2011, Montrul 2006), the syntactic-semantic distinction of between unaccusative and unergative verbs. In the present volume, Bonfatti-Sabbioni (2022) contributes new evidence from auxiliary selection showing robust knowledge of unaccusativity in Italian heritage speakers in the USA. Other area of target-like ability is syntactic and morphological processing, as in the comprehension of relative clauses in Spanish with syntactic ambiguity (Alguien disparó contra el criado de la actriz que estaba en el balcón con su marido. ‘Somebody shot the servant of the actress who was in the balcony with her husband.’), and the processing of inflected words. Spanish is a high attachment language (the servant was shot), while English is a low attachment language (the actress was shot). In a task testing comprehension of who was doing what, Keating et al. (2011) found that native Spanish speakers and heritage speakers preferred high attachment. Jegerski et al. (2016) used an online measure with monolingual native speakers in Mexico and heritage speakers of Mexican descent and found no differences between the native speakers and the heritage speakers in accuracy and reaction times: both groups preferred high attachment. In morphological processing, while L2 learners have been found not to analyze (i.e., decompose) morphology like native speakers (Clahsen et al. 2010), heritage speakers of Turkish (Uygun and Clahsen 2021) and of Spanish (Mason 2019) seem to process complex words like native speakers. Therefore, in some aspects of syntax, syntax-semantics and morphology as well as in syntactic and morphological processing heritage grammars converge with baseline grammars. Most noticeable, however, and as such the focus of intense research in the past three decades is that heritage languages diverge from baseline grammars in many more areas, especially in production of inflectional morphology, the syntax-discourse interface, some areas of semantics, and long-distance dependencies (agreement, binding, relative clauses) (Polinsky 2008). Divergence takes many forms, but it is usually described as some sort of simplification that can be related or unrelated to dominant language transfer or to lower proficiency in the heritage and concomitant processing limitations. Indeed, many of the linguistic patterns exhibited in heritage languages are neither random nor unusual, and point toward the universality of the underlying cognitive and linguistic processes that shape these grammars (Montrul 2016, Polinsky 2018).

10 

 Silvina Montrul

For example, heritage speakers of diverse languages show strikingly similar patterns of omission of obligatory inflectional morphology and regularization of irregular forms, processes leading to overall simplification of inflectional morphology in their heritage languages. These tendencies are observed both with nominal morphology (marking definiteness, case, gender agreement, plurality) and with verbal morphology (finiteness, agreement, tense, aspect, mood and evidentiality) (see Montrul 2016, Polinsky 2018 and Putnam et al. 2021 for overviews). In general, heritage speakers develop and retain solid knowledge of agreement and tense, but the inflectional morphology of categories that interface with semantics and pragmatics (aspect, mood, evidentiality) are more prone to simplification even in heritage speakers who have advanced proficiency and are quite fluent (Arslan et al. 2015, Montrul and Perpiñán 2011), and more so if the forms must be expressed in complex syntax, like subjunctive (Lynch 1999, Polinsky 2006, Silva-Corvalán 1994) and conditional verbs in adverbial subordinate clauses (Fairclough 2005). Nominal morphology seems to be more affected than verbal morphology (Benmamoun et al. 2013, Montrul 2016). Gender, number, and case in noun phrases are mastered at an early age (3 to 4 years old) by monolingual children, especially in languages that have relatively rich morphology such as Spanish (Montrul 2004), Italian (Romano 2022), Hungarian (Gábor and Lukács 2012) and Russian (Rodina and Westergaard 2012). Monolingual Russian and Spanish-speaking children control gender marking by age 3 or 4 with almost 95–100% accuracy, except for most irregular, less frequent, and marked forms. However, several studies of child and adult heritage speakers of different proficiency levels show that gender assignment, more than gender agreement, remains a problem area in heritage language grammars (Håkansson 1995, Lohndal and Westergaard 2016, Montrul and Potowski 2007, Montrul et al. 2008, Polinsky 2018), likely related to heritage speakers’ smaller lexicons and slower lexical retrieval and processing speed in their weaker language (Montrul et al. 2013). Erosion of case marking has been found in several heritage languages including Dyirbal (Schmidt 1985), Estonian (Maandi 1989), Pennsylvania German (Huffines  1989), Hindi (Montrul et al. 2012), Finnish (Larmouth 1974), Spanish (Montrul 2004), Russian (Polinsky 2006), and Korean (Kim et al. 2018). These languages vary significantly in their case typology and in the number of cases they mark overtly. Yet, the general pattern observed in the grammars of heritage speakers is that of reduction of case systems and simplification of morphological paradigms (regularization). For example, Russian has a six-way distinction in nouns: nominative, accusative, dative, instrumental, oblique, and genitive. This case system is severely reduced in heritage speakers (Polinsky 2006, 2008): dative is replaced by accusative and accusative by nominative in many constructions with subjects, direct,

Chapter 1 Heritage language development 

 11

and indirect objects. Thus, while homeland and adult immigrant speakers of Russian use the six-case markings, heritage speakers tend to use only two: nominative and accusative. Ergative-absolutive languages (Hindi, Greenlandic, Dyirbal, and Basque, among others) mark the subject of transitive verbs overtly with ergative case. Subjects of intransitive predicates and objects are marked with absolutive case, which is typically null (Butt 2006). Heritage speakers of ergative languages like Hindi (Montrul et al. 2012), Basque (Austin 2007), Inuktitut (Murasugi 2012), and Dyirbal (Schmidt 1985) frequently omit or no longer use ergative case marking. Another area of morphological reduction for heritage speakers is Differential Object Marking (DOM). DOM is the overt morphological marking of direct objects that are prominent on semantic and pragmatic scales (Aissen 2003). The general parameters that define DOM cross linguistically are animacy and referentiality (definiteness and specificity) of the object, the lexical semantics of the verb, and topicality (Torrego 1998, Von Heusinger and Kaiser 2005), among others, but the relative relevance of these parameters and the morphological expression of DOM differ from language to language. For example, in Spanish and Romanian, human and specific direct objects are marked with the prepositions “a” (Spanish) and “pe” (Romanian), whereas inanimate objects are typically unmarked. In Turkish and in Hindi, specific direct objects (mostly animate objects but also some inanimate objects), are also obligatorily marked with accusative case (Hindi –ko and Turkish  -yI). Research on the acquisition of DOM in languages like Korean, Lithuanian, Estonian, Romanian, Turkish, Spanish, and Farsi by young children indicates that DOM is acquired and mastered in oral production with close to 90% accuracy by age 3;00, at least when it comes to its morphological expression (Mardale and Montrul 2020). However, DOM is particularly vulnerable to morphological variability and omission in child and adult heritage language grammars when the majority language does not have DOM (Coşkun-Kunduz and Montrul 2022, Cuza et al. 2019, Montrul et al. 2015, Montrul and Bowles 2009, Montrul and Sánchez-Walker 2013, Ticio 2015). Compared to inflectional morphology, syntax is overall better preserved in heritage language grammars. However, heritage speakers tend to prefer SVO even when their language allows different options. In Spanish, Arabic, Italian and Greek, postverbal subjects are common in simple sentences and with particular verbs (i.e., unaccusatives), although they have a different pragmatic function from preverbal subjects. Several studies of child and adult heritage speakers of Spanish, Arabic, Greek and Italian have found that heritage speakers tend to favor preverbal subjects, especially when the majority language is a SVO language, like English (Albirini et al. 2011, Daskalaki et al. 2019, De Prada Pérez and Pascual y Cabo 2012, Hoot 2017, Listanti and Torregrossa 2021, Montrul 2004, Silva-Corvalán 1994), inde-

12 

 Silvina Montrul

pendently of discourse function. In some cases, preference for some particular word orders can be related to transfer from the majority language. Torregrossa and Listanti (2022) considered that some of the word order patterns observed in their Italian heritage children in Germany were due to transfer from German, the majority language. Another example comes from relative clauses, which show different word order patterns within the relative clause. Polinsky (2011) reported that heritage speakers of Russian in the United States made errors comprehending relative clause when the word order within the relative clause was different from SVO. Similarly, Sanchez-Walker (2019) found that the Spanish heritage in the United States speakers were very accurate with subject and object relative clauses when these had English word order (SVO), yet more inaccurate interpreting the sentences with subject-verb and object-verb inversion within the relative clause possible in Spanish. Another extensively investigated area of divergence for heritage speakers is the expression of null and overt subject pronouns in null subject languages. Numerous studies of heritage speakers of Italian (Serratrice et al. 2004), Spanish (Montrul and Sánchez-Walker 2015), Russian (Ivanova-Sullivan 2014), Greek (Kaltsa et al. 2015), among many others, found that heritage speakers tend to overuse overt subject pronouns in production in contexts where null subjects are pragmatically felicitous, such as in topic continuity contexts. This is what Laleko and Polinsky (2017) call the ‘silent problem’, or difficulty tracking or referencing null elements. In anaphoric contexts, heritage speakers interpret null and overt subjects as referential with subjects of the main clause, whereas native speakers tend to link null subjects to subjects but overt pronouns to the object antecedents. Redundant use of overt pronouns has received several explanations, ranging from transfer from the nonnull subject majority or contact language, cognitive complexity of processing in bilinguals, and erosion of the [+topic shift] pragmatic feature of overt pronouns (simplification), among others. These are just some representative examples of areas where heritage speakers exhibit divergence with respect to baseline grammars. What is important to note here is that despite having data from different languages, different speakers, different contact languages, and different methodologies, there are many strikingly common patterns of phonological, morphological and syntactic acquisition that stress the universality of the linguistic mechanisms involved. Among these patterns, Polinsky (2018) noted a propensity for-one-to one mappings, preference for acoustically salient, perceptual material, morphological and syntactic restructuring, and preference for default settings. To this we must add that age of onset of bilingualism and reduced exposure to the heritage language plays a significant role in heritage language development and ultimate attainment (Montrul 2008). Torregrossa and Listanti (2022) show that Italian heritage children in Germany

Chapter 1 Heritage language development 

 13

display different levels of acquisition of complex syntactic structures. The timing and extent of acquisition of these structures are sensitive to different language exposure variables as related to different age ranges. However, most studies with heritage language speakers look at heritage language development in a specific country or sociopolitical context, and in most studies the contact, majority and typically dominant language is English. As a result, it is difficult to tease apart the influence of the majority language from simplification that is not related to the majority language (Polinsky and Scontras 2019). The studies in this volume are unique because they are all about Italian as a heritage language, and include research conducted in different majority language contexts, which is a much-needed comparative dimension to advance our understanding of the variables that affect heritage language development, maintenance and loss. To this day, the role of the majority, dominant language in contributing to the patterns of divergent attainment and simplification in selected aspects of heritage speakers’ grammars remains poorly understood. This is because majority language transfer has not taken a prominent role in heritage language studies as it has in second or third language studies (Rothman et al. 2019).

3 Majority language transfer In many heritage language studies, allusion to transfer is typically indirect or even downplayed (Silva-Corvalán 1994). Polinsky (2018: 358) writes: “A large body of work on heritage languages has concerned itself with identifying the role of some of these factors and teasing apart their relative contributions. Progress has been made in this direction, and we are already in a position where we can offer some considerations. In particular, it appears that transfer from the dominant language is not as significant as one may think and less prominent than it is in second language acquisition” (emphasis mine). Romano’s (2021) study of knowledge of clitic constructions in very advanced heritage speakers of Italian and L2 speakers support Polinsky’s assertion. However, Romano (2022) shows that having gender in majority language Swedish facilitates the retention of gender agreement in Italian heritage speakers in Sweden. Like Romano (2022), I have shown that some of the transfer effects found in heritage language speakers are very similar to the patterns found in second language learners (Montrul 2010). The clearest example of this claim are the findings of my study with Tania Ionin on article semantics (Montrul and Ionin 2010, 2012). On the other hand, recent transnational studies of the same heritage language in different sociopolitical territories or countries have produced mixed results.

14 

 Silvina Montrul

3.1 Studies of heritage speakers and second language learners Montrul and Ionin (2012) investigated the interpretation of definite plural articles in Spanish by L2 learners of Spanish and Spanish heritage speakers whose dominant language is English. Although both Spanish and English have definite and indefinite articles, the languages vary in the semantic interpretations of these. For example, genericity in English is expressed through bare plural noun phrases, as in (1a). With the definite article, as in (1b), the sentence refers to a specific group of tigers. In Spanish, bare plurals in subject position are typically ungrammatical, as in (2a), but the definite article can be used to express both a generic statement and a specific statement. So, sentence (2b) can be a generic statement about tigers (generic reading) or can also refer to a specific group of tigers (specific reading), depending on the context. (1) a. Tigers eat meat. b. The tigers eat meat. (2) a.

generic specific

Tigres comen carne. ungrammatical Tigers eat meat ‘Tigers eat meat.’ b. Los tigres comen carne. generic, specific the tigers eat meat ‘The tigers eat meat.’ ✶

Montrul and Ionin (2012) asked whether L2 learners of Spanish and heritage speakers would tend to interpret definite plural determiners as generic, very much like Spanish native speakers tend to interpret definite articles in Spanish, or as specific due to transfer from English. Results of an acceptability judgment task and a truth value judgment task administered to native speakers of Spanish, Spanish heritage speakers, and L2 learners of Spanish showed significant differences between the native speakers and the two experimental groups. The native speakers preferred a generic interpretation for plural definite articles while the L2 learners and the heritage speakers showed a preference for specific readings instead. Although Montrul and Ionin set up their hypothesis to examine the effects of age of acquisition for transfer, there were no advantages for the heritage speakers in terms of more native-like knowledge of semantics, and the two groups showed the same patterns of dominant language transfer effects. Montrul and Ionin (2010) tested some of the same heritage speakers in English and showed that the heritage speakers had native interpretations of definite determiners in English. Therefore, these participants transferred the semantic interpretations from the stronger language onto the heritage language.

Chapter 1 Heritage language development 

 15

Kupisch (2012) reported similar results with Italian heritage speakers in Germany and German-speaking L2 learners of Italian. Like English and Spanish, German and Italian have definite and indefinite articles as well as contexts in which nouns can occur with or without articles. In both languages, bare noun phrases can occur in specific syntactic configurations, as shown in (3). (3) a. Jeden Tag isst sie Kartoffeln German today eats she potatoes b. Ogni giorno mangia patate. Italian every day eats potato “She eats potatoes every day.” (4) a. Katzen sind intelligent. Cats are intelligent b. ✶Gatti sono intelligenti. cats are intelligent c. I gatti sono intelligenti. the cats are intelligent “Cats are intelligent.”

German Italian

In Italian and German (and in English and Spanish), bare noun phrases can appear in object position, as in (3b). Bare nouns in German, as in English, have a generic meaning. In Italian (as in Spanish), bare plural nouns in subject position are ungrammatical (4b), and generic meanings are expressed with the definite article (4c). The participants were young adults recruited in Germany and Italy and came from binational families (German-Italian parents). The Italian heritage speakers were simultaneous bilinguals but some were dominant in Italian and others in German at the time of testing. The other group consisted of advanced late L2 learners of Italian, with German as their L1. The results of an acceptability judgment found significant differences in ratings on ungrammatical sentences with bare plurals in subject position by group. The heritage speakers who were dominant in Italian performed at ceiling like the Italian native speakers, but the heritage speakers with Italian as weaker language and the German-speaking L2 learners were only 33% and 56% accurate, respectively, with bare NPs in specific and generic contexts, as in (4c). The results of generic interpretations on sentences with definite plural articles tested in a truth value judgment task showed that the three groups imposed the interpretations of German onto Italian, showing preference for the specific more than the generic interpretation of definite articles in Italian. However, the heritage speakers who were dominant in Italian gave significantly fewer specific responses (and more generic responses) than the L2 learners of Italian. At

16 

 Silvina Montrul

least for the domain of article semantics, Kupisch (2012) agrees with Montrul and Ionin’s conclusion (2012), that when comparing L2 learners and heritage speakers, age of acquisition is overruled in this case by dominant language transfer. Transfer from the stronger language, in addition to reduced input conditions, plays a significant role in shaping adult heritage language grammars, as it does in L2 grammars. Therefore, studies that compared heritage speakers and second language learners have found that in some areas, the two groups show similar linguistic patterns related to transfer from the majority, dominant language.

3.2 Studies of heritage languages cross-linguistically and transnationally In studies that do not compare second language learners and heritage speakers, there are two ways to assess the potential role of dominant language transfer. One research design is cross-linguistic and involves identifying several languages that share a particular grammatical phenomenon not shared with the majority language. Several studied have examined the erosion of Differential Object Marking. Montrul, Bhatt and Girju (2015) and Montrul, (2022) used the same methodology to study DOM in Spanish, Hindi and Romanian as heritage languages in the United States. Two other examples are Chung’s (2018) study of DOM in Korean heritage speakers in the United States and Coşkun-Kunduz and Montrul’s (2022) study of Turkish heritage speakers in the United States. In all these studies the majority language is English, which does not exhibit DOM. The main findings of Montrul, Bhatt and Girju (2015) were that DOM was vulnerable to omission in required contexts in the three heritage languages, although to different degrees. Chung (2018) and Coşkun-Kunduz and Montrul (2022) confirmed the same trend: heritage speakers of Korean and of Turkish in the United States often omit DOM in required contexts. These studies and others highlight how particular grammatical phenomena are vulnerable in general, on the one hand, and how the patterns of erosion and simplification in this case may be related to influence from the majority language because English objects are unmarked. This is yet another example where it is hard to tease apart morphological simplification from majority language transfer. More evidence that DOM omission in Spanish heritage speakers is to a large extent related to the majority language comes from studies of heritage speakers in France (Grosjean and Py 1991) and Spanish as a heritage language in the Netherlands (Irizarri van Suchtelen 2016). French and Dutch do not have DOM. These heritage speakers of Spanish are also reported to omit DOM in their Spanish in oral production and grammaticality judgment tasks. Romanian heritage speakers in Italy have been reported to omit DOM (Cohal 2014). Finally, the reverse

Chapter 1 Heritage language development 

 17

is also attested: cases where the majority language has DOM and the minority language does not but develops DOM by contact with the majority language. Guaraní speakers in Paraguay and Basque speakers in Spain produce DOM in Guaraní and Basque, which is not attested in standard Guaraní and standard Basque (Rodríguez Ordóñez 2017, Shain and Tonhauser 2010). The other research design that could also be informative about majority language transfer is transnationally, comparing the same heritage language in contact with different majority languages. In addition to the findings with DOM mentioned earlier, several studies of Spanish as a heritage language in the United States have found instability with gender agreement in noun phrases, higher rates of overt subjects in contexts where null subjects are felicitous, variability with the use of the preterite and imperfect past tense forms, and replacement of indicative for subjunctive in subordinate clauses. The studies of Chilean heritage speakers in the Netherlands (Frasson et al. 2021, Irizarri van Suchtelen 2016, van Osch et al. 2014) found that the heritage speakers also produced errors of gender agreement, differential object marking, and variable uses of subjunctive. Another example of majority language transfer comes from Donoso (2016), who found that Chilean heritage speakers of Spanish in Sweden express motion events in Spanish as in Swedish instead of using the typical Spanish syntactic pattern. However, Schmitz et al. (2016) examined the production of null and overt subjects in Spanish and Italian heritage speakers in Germany and did not find overproduction of overt subjects in these groups as has been found for Spanish in the United States. Studies of Greek as a heritage language in the United Kingdom (Argyri and Sorace 2007), in Sweden (Kaltsa et al. 2015), in Canada (Daskalaki et al. 2020, Daskalaki et al. 2019) and in Chile (Giannakou 2018) are also informative. Most of these studies have investigated the expression of subjects, both the overt/null subject distribution and subject-verb word order as a function of information structure. The studies of Greek in contact with non-null subject languages (English, Swedish) found preference for SV word order irrespective of pragmatic effects and overproduction of overt subject in null subject contexts, which are typically attributed to difficulty integrating linguistic phenomena at the syntax-discourse interface (Sorace 2011). The study of Greek in contact with Spanish (Giannakou 2018), by contrast, found no attrition or inconsistent use of null subjects in Greek by first-generation immigrants and heritage speakers of Greek in Chile. Giannakou’s findings are consistent with positive transfer from the majority language Spanish, a null subject language. In a similar vein, Guijarro Fuentes et al. (2022) show no effect of Spanish on first generation Italian-speaking immigrants in Spain in the domain of clitic expression. Perhaps, when the minority and majority language are very typologically similar, the likelihood of the minority language developing innovative features is less likely (see D’Alessandro 2021).

18 

 Silvina Montrul

The cumulative knowledge we gain from individual studies on heritage languages in different contexts are important to understand the influence of specific factors, in this case dominant language transfer, on the linguistic patterns observed in heritage speakers of the same language. However, when contradictory findings are reported from individual studies, it is difficult to know whether these are related to the use of different methodologies by different studies testing the same grammatical phenomenon, or to the characteristics of the heritage community in their specific sociolinguistic context. For direct comparability, more studies of a given heritage language in different language-contact situations, using the same methodology are called for and, at present, there are very few. Rinke and Flores (2018) investigated the interpretation of overt and null subject pronouns in anaphoric contexts in heritage speakers of European Portuguese in Germany and Andorra with German and Spanish/Catalan as majority languages. The methodology was an offline sentence interpretation task testing sentences with null or over pronouns in adverbial subordinate clauses such as O avô fotografou o menino quando ele/pro saiu da garagem. ‘The grandfather took a picture of the boy when he left the garage.’ Who left the garage? The boy or the grandfather? In such ambiguous contexts, monolingual native speakers of null subject languages preferentially interpret the null subject in the subordinate clause as the subject referent of the main clause (the grandfather, which implies topic continuity) and the overt subject in the subordinate clause (ele) as being co-referential with the object of the main clause (the boy, signaling topic shift). These referential interpretations are captured by Carminati (2002)’s Position of Antecedent Hypothesis (PAH). The PAH predicts that, in ambiguous contexts, the null pronoun refers to an antecedent that is in the Inflectional Phrase (IP), subject-related position, whereas the overt pronoun tends to select an antecedent lower in the phrase structure, typically a non-subject antecedent. The PAH has been attested in several null subject languages, and cross-linguistic differences have been found with respect to the scope of overt subject pronouns (Filiaci et al. 2014, Romano 2019). Rinke and Flores (2018)’s results showed that, overall, monolingual controls and heritage speakers know that null pronouns are preferentially interpreted in terms of topic continuity and that overt subjects are related to topic shift. They are also sensitive to the syntactic context (intrasentential vs. intersentential) and the directionality of the anaphoric relation (anaphoric vs. cataphoric), although to different degrees. Whereas the monolingual groups performed as expected, the heritage speakers accepted more overt subject pronouns coreferential with objects than the monolingual speakers. With respect to transfer from the majority language, the comparison between the heritage speakers in Germany and the heritage speakers in Andorra revealed no statistical differences and shows that whether the contact language is a null subject language or not does not impact the European Portuguese heritage speakers’ knowl-

Chapter 1 Heritage language development 

 19

edge of anaphora resolution. Rinke and Flores (2018)’s results support Sorace et al. (2009)’s processing/cognitive resources hypothesis, and suggest that preference for overt subjects in topic continuity contexts is likely due to processing or cognitive resource limitations related to bilingualism as such. Most recently, Ionin et al. (2021) found suggestive evidence of majority language transfer in heritage speakers of Russian in the United States, in Israel and in Finland. Ionin et al. examined whether heritage speakers of Russian acquire the relationship between word order (SVO vs. OVS) and information status (old vs. new) in simple sentences. The study tested non-emotive sentences with neutral intonation with SVO order and OVS (scrambling). Participants were asked who did what: Who did the dog bite? (object) or Who bit the dog? (subject). A response with SVO (The dog bit a cat) is appropriate if the object is new information and OSV (A cat bit the dog) if the subject is new information. The research questions were whether heritage speakers would substitute (in)definiteness for information status, due to transfer from their dominant language, and whether Russian heritage speakers with different dominant languages (English, Hebrew and Finnish) would exhibit different overall patterns. English is an SVO language, while Hebrew and Finnish allow different word orders and have overt case marking, like Russian. Finnish has scrambling like Russian while Hebrew does not. The main task used was an acceptability judgment task (AJT). Short dialogues were presented in both auditory and written form with a sentence underneath to rate the acceptability of the answer to the questions in the dialogue. According to the results, heritage speakers of Russian recognize that word order (SVO vs. OVS) is related to information status (old vs. new), but the patterns were clearer with object questions (SVO preferred) than with subject questions (OVS preferred). They did not incorrectly map word order to (in)definiteness rather than information structure, showing the same preferences and patterns of performance with common as with proper nouns, even though (in)definiteness is related to word order only with the former. However, heritage speakers of Russian with different dominant languages (English vs. Hebrew vs. Finnish) behaved differently. The Finnish-dominant heritage speakers were the most target-like, while the English-dominant ones were the least target-like. The Hebrew-dominant heritage speakers were more accurate than the English-dominant heritage speakers, but when proficiency was factored into the analysis, there was no difference between the Hebrew and English-dominant speakers. The sample size for the Finnish-dominant speakers was also very small, compared to the other two groups. The advantage for the Finnish-dominant heritage speakers could be due to several factors, as Ionin et al. (2021) mention. First, the Finnish-dominant speakers were the most proficient in Russian, and the only ones whose dominant language has scrambling. Thus, both proficiency and transfer are possible explanations. Although the Hebrew-dominant speakers were not much

20 

 Silvina Montrul

lower than Finnish-dominant ones in terms of proficiency, they were much less target-like on word order. So, transfer from Finnish is probably the right explanation, but the results are inconclusive because when the Hebrew and English groups were compared with proficiency controlled in the statistical analysis, the language effect disappeared. Ionin et al. (2021) also tested English-speaking L2 learners of Russian with the same task. They had overall lower proficiency in Russian, and did much worse than the Russian heritage speakers on the AJT: both word orders were allowed, with a preference for SVO regardless of information structure. Compared to the performance of English-speaking L2 learners of Russian, Ionin et al.’s findings suggest that the word order-information structure relationship is relatively robust in Russian heritage speakers; unlike morphosyntax, which has been found to be incompletely acquired or attrited (Polinsky 2006). In summary, while the dominant language plays a role in the linguistic patterns observed in heritage language grammars, its effects are not always evident. So far, the strongest evidence comes from studies comparing heritage speakers and L2 learners. When heritage speakers of the same heritage languages in contact with other majority languages are compared, there are other factors that confound the findings, such as proficiency, which is in turn related to other sociolinguistic and sociopolitical factors.

4 The sociopolitical dimension When conducting comparative studies of a given heritage language in different countries, a factor that can play a role and which has rarely been discussed in linguistic studies of heritage languages is the sociopolitical contexts. Sociolinguistic studies do examine this dimension but often do so in general and not in relation to specific grammatical features. For example, in their study of Spanish as a heritage language in Israel, Stavans and Ashkenazi (2022) focus more on family language practices and the role of schooling across generations than on specific linguistic features affected by contact with Hebrew in this situation. Although relating performance on specific linguistic features to the status and size of the heritage language in the host society seems like a stretch, it may in fact not be so if we consider the many factors that determine the linguistic proficiency of heritage speakers, as represented in Figure 1. Most linguistic studies focus on the innermost little circle, such as the acquisition or present knowledge of specific areas of grammatical knowledge, such as DOM, gender agreement, word order, perception of lexical stress, and so on. For many areas of linguistic knowledge, there is a relationship between the proficiency of the

Chapter 1 Heritage language development 

 21

Figure 1: External factors affecting specific linguistic features in heritage language competence and use.

heritage speakers in the heritage language and the degree of acquisition of the grammatical structure of interest (next innermost circle, grammatical and communicative competence). A good example is Polinsky (2006)’s study of heritage Russian who shows these correlations. Key questions that have also taken center stage in the study of heritage grammars are the main source of variability in proficiency levels and the degree of grammar simplification observed, such as the quantity and the quality of the input that heritage speakers are exposed to during childhood and after the onset of schooling (which is usually in the majority language). The middle layer in Figure 1 is language practices (input and use). The amount and quality of input interact with the heritage speakers’ age. I have argued that insufficient input and use of the heritage language during the critical period for language learning leads to partial or incomplete acquisition and or/attrition later in childhood of aspects of language that depend more on extensive exposure and language use for linguistic consolidation and maintenance (Montrul 2008, 2016). Heritage speakers appear to challenge the hypothesis that early onset of acquisition is all that is needed for native-like ultimate attainment in adulthood. As shown, many heritage speakers do not reach native-like ability in all aspects and domains of their heritage language in late childhood and beyond, showing divergent patterns compared to baseline grammars, including transfer effects and simplification, as discussed earlier. Many heritage speakers do eventually reach nativelike ability

22 

 Silvina Montrul

in most aspects of the dominant majority language, which in some cases is their second language acquired after age four (Lloyd-Smith et al. 2020). Other external situational factors also condition input and use of the heritage language and contribute to shape the languages, perhaps more indirectly. The outer layer in Figure 1 is the sociopolitical context, which appears to be quite removed from how a specific linguistic feature may be affected in heritage language grammars. While the sociopolitical environment does not directly determine why some heritage speakers, say, produce gender agreement consistently whereas others show high variability and indeterminacy, it can play a role indirectly, in the following way. Attitudes, beliefs, and levels of education is the layer that relates the outer layer with the inner layers. The status of the heritage language in the host society determines the degree of vitality and visibility of the heritage language beyond the home, language attitudes toward the heritage language by its speakers and by the members of the majority society, and the availability of the heritage language in the education system. Input and use of the heritage language are embedded in attitudes and beliefs about the heritage language by heritage speakers themselves and the value the heritage language community vests on preserving and transmitting their heritage language. Language attitudes and desires to maintain and transmit the language to future generations drive communicative practices and language choices in heritage language families. These practices directly determine the quantity and quality of input that heritage speakers are exposed to at home or beyond the home, as well as their active use of the language. For example, Tyrell et al. (2014) discuss how the practices of Spanish language use by Spanish-speaking families in Britain are rooted in their sense of identity and belonging. While for the parents their use of Spanish is tied to their identities as Spanish-speakers, for their children using Spanish and English are the pillars of their bilingual identity. An example of the potential role of attitude and identity on grammatical features is offered by Albirini et al. (2011). They found that heritage speakers of Palestinian Arabic in the United States had higher accuracy on several morphosyntactic aspects of Arabic (word order, null subjects, prepositions, plural morphology, possession) than Egyptian Arabic heritage speakers in the United States. While many of the non-convergent patterns produced by the heritage speakers could be related to influence from English, the majority language, this would apply to the two groups. But why would transfer from English affect one Arabic nationality more than the other? The authors speculated that a possible reason for the difference may have to do with identity and attitudes. Palestinian parents might vest the Arabic language with a greater sense of identity than the Egyptian parents, possibly because of the geopolitical context from which the Palestinian parents came. Language may be one of the strongest links to their Palestinian and Arab roots in a world in which many Palestinians feel a strong need to maintain those ties. However, because this study

Chapter 1 Heritage language development 

 23

did not include an extensive sociolinguistic interview to assess this possibility, this explanation, although likely, can only be tentative. Thus, the influence of situational factors in cases where the dominant language exerts significant influence on the heritage language remains to be understood and calls for research integrating methodologies from psycholinguistics and sociolinguistics. The availability and vitality of the heritage language beyond the home in a given society also plays a role. This of course impacts education in the heritage language: whether it is available in the school system, through community schools, or not at all. The degree to which communities and societies provide education in the heritage sends specific messages about the status and importance of the heritage language. Ionin et al. (2021) note that the Russian-speaking community in Israel is quite large, with more than a million and a half speakers representing 17.5% of the population of Israel. Not only are these speakers highly-educated but they also have a strong commitment to retaining Russian, and promoting mother tongue instruction in Russian through Russian literacy schools (Schwartz 2008). In Finland, there are between 40,000 and 60,000 Russian speakers, constituting the largest the immigrant language in Finland. Education in Russian is available through mother tongue instruction in schools as well as through preschools and day care centers. Protassova (2008) reports that in 2004 half the total number of Russian-speaking school children received instruction in Russian as a heritage language. The Russians speakers in the United States are also highly educated (Kagan and Dillon 2008), but the teaching of Russian beyond the home is less available, perhaps, than in Israel and Finland. Schooling in the heritage language plays a substantial role in language maintenance and in lifting the status of the language among its speakers. A follow-up to the Ionin et al. (2021) study mentioned earlier would be to examine whether knowledge of word order in heritage Russian is related to the type and amount of education that heritage speakers receive in Russian. Rodina et al. (2020)’s study of Russian-speaking heritage language children (ages five to seven) in different countries addresses the importance of input and education as a function of different national contexts. Rodina et al. (2020) considered possible effects of majority language transfer together with language-external variables, such as language exposure and use, literacy training, and sociodemographic origin of the speakers. The heritage language studied was Russian and the focus was sensitivity to morphophonological cues on gender agreement in nominative singular noun endings. The majority languages were Norwegian, German, Hebrew, Latvian, and English, which show similarities and differences in gender assignment and provided an excellent test case for assessing the effects of majority language transfer on child bilingual acquisition of grammatical gender. In addition, Rodina et al. considered the extent to which Russian bilingual children used morphophonological gender cues in the context of heritage language instruction in

24 

 Silvina Montrul

Norway, Germany, Israel, Latvia, and the United Kingdom that have very active heritage Russian communities. The children had received Russian language and literacy exposure in complementary schools and through Russian immersion programs. Gender agreement was elicited via two oral production tasks, one with real Russian nouns and the other with novel nouns, testing feminine, masculine and neuter word forms. The overall results showed that whether the contact, majority language had or did not have gender or expressed it differently did not affect (positively or negatively) the bilingual children’s sensitivity to gender cues in their heritage language. The gender systems exhibiting high transparency and predictability (like Latvian and Hebrew) did not seem to lead to an increased awareness of the Russian gender system. And the systems exhibiting reduced transparency and predictability (like German and Norwegian) did not cause delays. Rodina et al. argue that the reason for this result is that the acquisition of gender in Russian is largely predicted by the gender cues available in Russian word endings and by other situational factors. One of the main predictors of gender assignment accuracy in the children was language exposure at home defined in terms of family type (one or two Russian-speaking parents). They also found that gender assignment accuracy was indirectly related to the size of the Russian immigrant community and especially to current exposure to and instruction in Russian as a heritage language. Overall, older children, children from communities with higher proportion of Russian speakers, as well as children receiving more exposure to Russian through instruction, acquired gender more easily than the children from communities with a lower percentage of Russian speakers and less exposure to Russian instruction. In other words, the less input the children have had in Russian (younger children, children from mixed families, and those who started kindergarten early), the more likely they were to develop a reduced gender system in Russian. Thus, the amount of input in Russian inside and outside the home is beneficial for heritage language acquisition and strengthens the likelihood that these speakers will show little simplification of the Russian gender system in their heritage language as adults. Rodina et al. (2020) aptly demonstrate how the sociopolitical context, and the status of the language within each context, determines many of the factors in the inner layers affecting language input and use (Figure 1), since the same heritage language can have very different fates in one sociopolitical context than in another. Up until now, the broader sociopolitical context has been overlooked in linguistic and psycholinguistic approaches to heritage grammars concerned with specific grammatical phenomena. There are differences between heritage languages spoken in different sociolinguistic and sociopolitical contexts that remain to be understood, as well as the role of the majority language in effecting many of the changes seen in heritage language speakers.

Chapter 1 Heritage language development 

 25

Other research designs that could shed more light on this issue would include a comparison of the same heritage language in different sociopolitical contexts keeping the majority language constant. An example would be a study of heritage speakers of Hindi in the United States and heritage speakers of Hindi in the United Kingdom. Hindi speakers in the United States are of very high socioeconomic status (SES) (Montrul 2022) whereas Hindi speakers in the United Kingdom are a more heterogeneous group including different castes, religions, education and levels of SES. To my knowledge, there are no linguistic or psycholinguistic studies of Hindi as a heritage language in the United Kingdom, even though reports of language proficiency in second and third generation heritage speakers suggest that the language is spoken mostly at home. Immigrants from India are the largest immigrant group in the United Kingdom, making about 25% of total immigrants (Barn 2008). Barn’s survey-based study found that the Hindi and Sikh families interviewed mostly speak English only at home or English and Hindi, but less than 20% spoke Hindi exclusively, even when they consider that maintaining and transmitting Hindi is important to them. Like most community language schools across the world, complementary schools in the United Kingdom are voluntary-run organizations that complement mainstream schooling by providing language-and literacy-focused teaching over the weekend or in the evenings. The main objectives of such schools are to promote community language(s) and literacy to children born into varied linguistic and cultural heritages (Creese 2009). These schools are available in the United Kingdom, but less so in the United States. If Rodina et al.’s study is instructive, we could investigate the extent to which the use of language in the home and the availability of Hindi instruction in the two contexts, in addition to other language identity and attitudinal variables, contribute to heritage language development, maintenance and ultimate attainment in adulthood. This is just an example of how to continue to make strides in identifying the many factors that play a role in heritage language development and maintenance.

5 Conclusion In the last two decades we have come to understand the grammatical and psycholinguistic development of heritage languages and we have uncovered many similarities about heritage languages. While much research has focused on understanding how exposure and input impact heritage language development, especially during childhood, we have found that reduced input affects different grammatical areas differently, and heritage speakers with lower proficiency tend to display divergent patterns and simplification. It is still not clear whether the patterns of simplification

26 

 Silvina Montrul

observed are general tendencies driven by limited language processing resources in the heritage language or by crosslinguistic influence or transfer from the dominant majority language. In many cases the dominant language has been English, and this language does not exhibit many grammatical or morphosyntactic properties of most of the heritage languages studied to date. Yet, the role of majority language transfer in heritage language knowledge is still poorly understood. In some grammatical areas, and in studies that have directly compared L2 learners and heritage language speakers, majority language transfer is related to the patterns of simplification found, whereas in other grammatical areas investigated in transnational studies that examine the same language in different sociopolitical contexts and in contact with different majority languages, the jury is still out. This is perhaps because in these situations, other external sociolinguistic and sociopolitical factors are confounded, and it is difficult to isolate the role of dominant language transfer and the external factors that also contribute to higher or lower proficiency in the heritage language. With their focus on Italian, many of the studies presented in this volume contribute to confirm findings from previous studies and offer new data and new perspectives on the role of the dominant language, proficiency and contextual factors in both heritage language development and L1 attrition. Future studies of specific languages that could isolate these important factors would allow us to understand their impact in heritage language development.

References Ahn, Sunyoung, Charles Chang, Robert DeKeyser & Sunyoung Lee‐Ellis. 2017. Age effects in first language attrition: Speech perception by Korean‐English bilinguals. Language Learning 67(3). 694–733. Aissen, Judith. 2003. Differential object marking: Iconicity vs. Economy. Natural Language & Linguistic Theory 21(3). 435–448. Albirini, Abdulkafi, Elabbas Benmamoun & Eman Saadah. 2011. Grammatical features of Egyptian and Palestinian Arabic heritage speakers’ oral production Studies in Second Language Acquisition 33(2). 273–303. Amengual, Mark. 2019. Type of early bilingualism and its effect on the acoustic realization of allophonic variants: Early sequential and simultaneous bilinguals. International Journal of Bilingualism 23(5). 954–970. Argyri, Ephrosyni & Antonella Sorace. 2007. Crosslinguistic influence and language dominance in older bilingual children. Bilingualism: Language and Cognition 10(1). 79–99. Arslan, Seçkin, Roelien Bastiaanse & Claudia Felser. 2015. Looking at the evidence in visual world: eye-movements reveal how bilingual and monolingual Turkish speakers process grammatical evidentiality. Frontiers in psychology 6. 1387. Austin, Jennifer. 2007. Grammatical interference and the acquisition of ergative case in bilingual children learning Basque and Spanish. Bilingualism: Language and cognition 10(3). 315–331.

Chapter 1 Heritage language development 

 27

Barn, Ravinder. 2008. Indian diaspora in the United Kingdom: Second-generation parent’s views and experiences on heritage language transmission. In Parvati Raghuram, Ajaya Kumar Sahoo, Brij Maharaj & Dave Sangha (eds.), Tracing an Indian Diaspora, 191–209. London: Sage. Benmamoun, Elabbas, Silvina Montrul & Maria Polinsky. 2013. Heritage languages and their speakers: Opportunities and challenges for linguistics. Theoretical Linguistics 39 (3–4). 129–181. Bonfatti-Sabbioni, Teresa. 2023 (this volume). Auxiliary selection in heritage speakers of Italian. In Francesco Romano (ed.), Studies in Italian as a Heritage Language. Berlin: De Gruyter Mouton. Butt, Miriam. 2006. Theories of Case. Cambridge: Cambridge University Press. Carminati, Maria Nella. 2002. The processing of Italian subject pronouns. Unpublished doctoral dissertation. University of Massachusetts Amherst. Chang, Charles. 2016. Bilingual perceptual benefits of experience with a heritage language. Bilingualism: Language and Cognition 19(4). 791–809. Chang, Charles, Yao Yao, Erin Haynes & Russell Rhodes. 2011. Production of phonetic and phonological contrast by heritage speakers of Mandarin. The Journal of the Acoustical Society of America 129(6). 3964–3980. Chung, Eun Seun. 2018. Second and heritage language acquisition of Korean case drop. Bilingualism: Language and cognition 21(1). 63–79. Clahsen, Harald, Claudia Felser, Kathleen Neubauer, Mikako Sato & Renita Silva. 2010. Morphological structure in native and nonnative language processing. Language Learning 60(1). 21–43. Cohal, Alexandru. 2014. Mutamenti nel romeno di immigrati in Italia. Milano: FrancoAngeli. Coşkun-Kunduz, Aylin & Silvina Montrul. 2022. Sources of variability in the acquisition of Differential Object Marking by Turkish heritage language children in the United States. Bilingualism: Language and Cognition 25(4). 603–616. Creese, Angela. 2009. Building on young people’s linguistic and cultural continuity: Complementary schools in the United Kingdom. Theory into Practice 48(4). 267–273. Cuza, Alejandro, Lauren Miller, Rocío Pérez Tattam & Mariluz Ortiz Vergara. 2019. Structure complexity effects in child heritage Spanish: The case of the Spanish personal a. International Journal of Bilingualism 23(6). 1333–1357. D’Alessandro, Roberta. 2021. Syntactic change in contact: Romance. Annual Review of Linguistics, 309–328. Daskalaki, Evangelia, Elma Blom, Vasiliki Chondrogianni & Johanne Paradis. 2020. Effects of parental input quality in child heritage language acquisition. Journal of Child Language 47(4). 1–28. Daskalaki, Evangelia, Vasiliki Chondrogianni, Elma Blom, Froso Argyri & Johanne Paradis. 2019. Input effects across domains: The case of Greek subjects in child heritage language. Second Language Research 35(3). 421–445. De Prada Pérez, Ana & Diego Pascual y Cabo. 2012. Interface heritage speech across proficiencies: unaccusativity, focus, and subject position in Spanish. In Kimberly Geeslin & Manuel Díaz-Campos (eds.), Selected Proceedings of the 14th Hispanic Linguistics Symposium, 308–318. Somerville, MA: Cascadilla Proceedings Project. Donoso, Alejandra. 2016. Expresiones de movimiento en español como segunda lengua y como lengua heredada: Conceptualización y entrega del Camino, la Manera y la Base [Movement expressions in Spanish as a second language and as a heritage language: Conceptualization and manifestation of path, manner and base], Stockholm University. Eide, Kristin & Arnstein Hjelde. 2015. Verb second and finiteness morphology in Norwegian heritage language of the American midwest. In Richard Page & Michael Putnam (eds.), Moribund Germanic Heritage Languages in North America, 64–101. Leiden: Brill.

28 

 Silvina Montrul

Fairclough, Marta. 2005. Spanish and heritage language education in the United States: Struggling with hypotheticals. Spanish and Heritage Language Education in the United States. Frankfurt am Main: Vervuert Verlagsgesellschaft. Filiaci, Francesca, Antonella Sorace & Manuel Carreiras. 2014. Anaphoric biases of null and overt subjects in Italian and Spanish: a cross-linguistic comparison. Language, Cognition and Neuroscience 29(7). 825–843. Frasson, Alberto, Roberta D’Alessandro & Brechje van Osch, B. 2021. Subject clitics in microcontact: A case study from heritage Friulian in Argentina and Brazil. Heritage Language Journal 18(1). 1–36. Fukuda, Shin. 2017. Floating numeral quantifiers as an unaccusative diagnostic in native, heritage, and L2 Japanese speakers. Language Acquisition 24(3). 169–208. Gábor, Bálint & Ágnes Lukács. 2012. Early morphological productivity in Hungarian: Evidence from sentence repetition and elicited production. Journal of Child Language 39(2). 411–442. Giannakou, Aretousa. 2018. Spanish and Greek subjects in contact: Greek as a heritage language in Chile. Unpublished doctoral dissertation. Cambridge University. Guijarro-Fuentes, Pedro, Iria Bello-Viruega, Estela García-Alcaraz & Sergio Viveros-Guzmán. 2023 (this volume). Ultimate attainment in long-immersed Italian immigrants: syntactic and semantic knowledge of direct object clitics and partitive ne. In Francesco Romano (ed.), Studies in Italian as a heritage language. Berlin: De Gruyter Mouton. Gor, Kira. 2014. Raspberry, not a car: context predictability and a phonological advantage in early and late learners’ processing of speech in noise. Frontiers in Psychology 5. 1449. Grosjean, François & Bernard Py. 1991. La restructuration d’une première langue: l’intégration de variantes de contact dans la compétence de migrants bilingues [Restructuring of a first language: The integration of contact variants in the competence of bilingual migrants]. La linguistique 27(2). 35–60. Håkansson, Gisela. 1995. Syntax and morphology in language attrition: A study of five bilingual expatriate Swedes. International Journal of Applied Linguistics 5(2), 153–169. Hoot, Brad. 2017. Narrow presentational focus in heritage Spanish and the syntax‒discourse interface. Linguistic Approaches to Bilingualism 7(1), 63–95. Huffines, Marion Lois. 1989. Case usage among the Pennsylvania German sectarians and nonsectarians. In Nancy Dorian (ed.), Investigating obsolescence: Studies in language contraction and death, 211–226. Cambridge: Cambridge University Press. Ionin, Tania, Maria Goldshtein, Tatiana Luchkin & Sofya Styrina. 2021. Who did what to whom, and what did we already know? Word order and information structure in heritage and L2 Russian. Linguistic Approaches to Bilingualism. https://doi.org/10.1075/lab.20045.ion Irizarri van Suchtelen, Pablo. 2016. Spanish as a heritage language in the Netherlands. A cognitive linguistic exploration. Utrecht: LOT. Ivanova-Sullivan, Tania. 2014. Interpretation of anaphoric subject pronouns in heritage Russian. In Ryan T. Miller Ryan T. Miller, Katherine I. Martin, Chelsea M. Eddington, Ashlie Henery, Nausica Marcos Miguel, Alison M. Tseng, Alba Tuninetti & Daniel Walter, Selected Proceedings of the 2012 Second Language Research Forum, 145–154. Somerville, MA: Cascadilla Proceedings. Project.Jegerski, Jill, Bill VanPatten & Gregory Keating. 2016. Relative clause attachment preferences in early and late Spanish-English bilinguals. In Diego Pascual y Cabo (ed.), Advances in Spanish as a Heritage Language, 81–99. Amsterdam: John Benjamins. Kagan, Olga & Kathleen Dillon. 2008. Issues in heritage language learning in the United States. In Nelleke Van Deusen-Scholl & Nancy Hornberger (eds.), Encyclopedia of Language and Education, 2nd Edition, Volume 4: Second and Foreign Language Education, 143–156. Springer Science+Business Media LLC.

Chapter 1 Heritage language development 

 29

Kaltsa, Maria, Ianthi Tsimpli & Jason Rothman. 2015. Exploring the source of differences and similarities in L1 attrition and heritage speaker competence: Evidence from pronominal resolution. Lingua 164, 266–288. Keating, Gregory, Bill VanPatten & Jill Jegerski. 2011. Who was walking on the beach? Anaphoraresolution in Spanish heritage speakers and adult second language learners. Studies in Second Language Acquisition 33, 193–221. Kim, Ji Young. 2020. Discrepancy between heritage speakers’ use of suprasegmental cues in the perception and production of Spanish lexical stress. Bilingualism: Language and cognition 23(2), 233–250. Kim, Kitaek, William O’Grady & Bonnie Schwartz. 2018. Case in heritage Korean. Linguistic Approaches to Bilingualism 8(2), 252–282. Knightly, Leah, Sun-Ah Jun, Janet Oh & Terry Kit-fong Au. 2003. Production benefits of childhood overhearing. The Journal of the Acoustical Society of America 114(1), 465–474. Kupisch, Tanja. 2012. Specific and generic subjects in the Italian of German-Italian simultaneous bilinguals and L2 learners. Bilingualism: Language and Cognition 15(4), 736–756. Kupisch, Tanja, Dagmar Barton, Katja Hailer, Ewgenia Klaschik, Ilse Stangen, Tatjana Lein & Joost van de Weijer. 2014. Foreign accent in adult simultaneous bilinguals. Heritage Language Journal 11(2), 123–150. Laleko, Oksana & Maria Polinsky. 2017. Silence is difficult: On missing elements in bilingual grammars. Zeitschrift für Sprachwissenschaft 36(1), 135–163. Larmouth, Donald Wilford 1974. Differential interference in American Finnish cases. Language 50(2), 356–366. Larsson, Ida & Janne Johannessen. 2015. Incomplete Acquisition and Verb Placement in Heritage Scandinavian. In Richard Page & Michael Putnam (eds.). Moribund Germanic heritage languages in North America, 153–189. Leiden: Brill. Lee, Teresa. 2011. Grammatical knowledge of Korean heritage speakers: Early vs. late bilinguals. Linguistic Approaches to Bilingualism 1(2), 149–174. Listanti, Andrea & Jacopo Torregrossa. 2021. The acquisition of post-verbal subjects in heritage Italian: How timing of L1–acquisition modulates the acquisition of syntax-discourse interface structures. RUEG Conference – Dynamics of Language Contact, Berlin, Germany, February. Lloyd-Smith, Anika, Marieke Einfeldt & Tanja Kupisch. 2020. Italian-German bilinguals: The effects of heritage language use on accent in early-acquired languages. International Journal of Bilingualism 24(2), 289–304. Lohndal, Terje & Marie Westergaard. 2016. Grammatical gender in American Norwegian heritage language: Stability or attrition? Frontiers in Psychology 7, 344. Lukyanchenko, Ana & Kira Gor. 2011. Perceptual correlates of phonological representations in heritage speakers and L2 learners. Proceedings of the 35th annual Boston University conference on language development, 414–426. Somerville, MA: Cascadilla Press. Lynch, Andrew. 1999. The subjunctive in Miami Cuban Spanish: Bilingualism, contact, and language variability. Unpublished doctoral dissertation. University of Minnesota. Maandi, Katrin. 1989. Estonian among immigrants in Sweden. In Nancy Dorian (ed.), Investigating obsolescence: Studies in language contraction and death, 227–241. Cambridge: Cambridge University Press. Mardale, Alexandru & Silvina Montrul. 2020. Differential object marking and its acquisition in different languages and contexts. In Alexandru Mardale & Silvina Montrul (eds.), The Acquisition of Differential Object Marking , 1–20. Amsterdam: John Benjamins.

30 

 Silvina Montrul

Mason, Sarah Ann. (2019). The influence of task factors and language background on morphological processing in Spanish. Unpublished doctoral dissertation, University of Illinois at UrbanaChampaign. Montrul, Silvina. 2006. On the bilingual competence of Spanish heritage speakers: Syntax, lexicalsemantics and processing. International Journal of Bilingualism 10(1), 37–69. Montrul, Silvina. 2008. Incomplete Acquisition in Bilingualism: Re-examining the Age Factor. Amsterdam: John Benjamins. Montrul, Silvina. 2010. How similar are adult second language learners and Spanish heritage speakers? Spanish clitics and word order. Applied Psycholinguistics 31(1), 167–207. Montrul, Silvina. 2016. The Acquisition of Heritage Languages. Cambridge: Cambridge University Press. Montrul, Silvina. 2022. Native Speakers, Interrupted. Cambridge: Cambridge University Press. Montrul, Silvina, Rakesh Bhatt & Roxana Girju. 2015. Differential object marking in Spanish, Hindi, and Romanian as heritage languages. Language 91(3), 564–610. Montrul, Silvina & Melissa Bowles. 2009. Back to basics: Differential object marking under incomplete acquisition in Spanish heritage speakers. Bilingualism 12(3), 363–383. Montrul, Silvina, Israel De La Fuente, Justin Davidson & Rebecca Foote. 2013. The role of experience in the acquisition and production of diminutives and gender in Spanish: Evidence from L2 learners and heritage speakers. Second Language Research 29(1), 87–118. Montrul, Silvina & Silvia Perpiñán. 2011. Assessing differences and similarities between instructed heritage language learners and L2 learners in their knowledge of Spanish tense-aspect and mood (TAM) morphology. Heritage Language Journal 8(1), 90–133. Montrul, Silvina & Kim Potowski. 2007. Command of gender agreement in school-age Spanish-English bilingual children. International Journal of Bilingualism 11(3), 301–328. Montrul, Silvina & Noelia Sánchez-Walker. 2013. Differential object marking in child and adult Spanish heritage speakers. Language Acquisition 20(2), 109–132. Montrul, Silvina & Noelia Sánchez-Walker. 2015. Subject expression in bilingual school-age children in the United States. In Ana Carvalho, Rafael Orozco & Naomi Shin (eds.), Spanish Subject Pronoun Expression in Spanish: A Cross-Dialectical Perspective, 231–247. Washington DC: Georgetown University Press. Montrul, Silvina. 2004. Subject and object expression in Spanish heritage speakers: A case of morpho-syntactic convergence. Bilingualism: Language and cognition 7(2), 1–18. Montrul, Silvina, Rakesh Bhatt & Archna Bhatia. 2012. Erosion of case and agreement in Hindi heritage speakers. Linguistic Approaches to Bilingualism 2(2), 141–176. Montrul, Silvina, Rebecca Foote & Silvia Perpiñán. 2008. Gender agreement in adult second language learners and Spanish heritage speakers: The effects of age and context of acquisition. Language Learning 58(3), 503–553. Montrul, Silvina & Tania Ionin. 2010. Transfer effects in the interpretation of definite articles by Spanish heritage speakers. Bilingualism: Language and Cognition 13(4) 449–473. Montrul, Silvina & Tania Ionin. 2012. Dominant language transfer in Spanish heritage speakers and L2 learners in the interpretation of definite articles Modern Language Journal 96(1) 70–94. Montrul, Silvina & Maria Polinsky. 2021. The Cambridge Handbook of Heritage Languages and Linguistics. Cambridge: Cambridge University Press. Murasugi, Keiko. 2012. Ergative and antipassive structures in bilingual Inuktitut speakers. Poster presented at the Conference on Formal Approaches to Heritage Language, University of Massachusetts, Amherst, April. Polinsky, Maria. 2006. Incomplete acquisition: American russian. Journal of Slavic Linguistics 14(2) 191–262.

Chapter 1 Heritage language development 

 31

Polinsky, Maria. 2008. Gender Under Incomplete Acquisition: Heritage Speakers’ Knowledge of Noun Categorization. Heritage Language Journal 6(1). DOI:10.46538/hlj.6.1.3 Polinsky, Maria. 2011. Reanalysis in adult heritage language: A case for attrition. Studies in Second Language Acquisition 33 (2), 305–328. Polinsky, Maria 2018. Heritage Languages and Their Speakers. Cambrideg: Cambridge University Press. Polinsky, Maria & Gregory Scontras. 2019. Understanding heritage languages. Bilingualism: Language and Cognition 22(5). 1–17. Protassova, Ekaterina. 2008. Teaching Russian as a heritage language in Finland. Heritage Language Journal 6(1). 127–152. Putnam, Michael, Lara Schwarz & Andrew Hoffman. 2021. Morphology of heritage languages. In Silvina Montrul & Maria Polinsky (eds.), The Cambridge Handbook of Heritage Languages and Linguistics, 613–643. Cambridge: Cambridge University Press. Rinke, Esther & Cristina Flores. 2018. Another look at the interpretation of overt and null pronominal subjects in bilingual language acquisition: Heritage Portuguese in contact with German and Spanish. Glossa: a journal of general linguistics 3(1) 68. 1–24. Rodina, Yulia, Tanja Kupisch, Natalia Meir, Natalia Mitrofanova, Olga Urek & Marit Westergaard. 2020. Internal and external factors in heritage language acquisition: evidence from heritage Russian in Israel, Germany, Norway, Latvia and the United Kingdom. Frontiers in Education, 5, DOI=10.3389/feduc.2020.00020 Rodina, Yulia & Marit Westergaard. 2012. A cue-based approach to the acquisition of grammatical gender in Russian. Journal of Child Language 39(5). 1077–1106. Rodríguez Ordóñez, Itxaso. 2017. Reexamining differential object marking as a linguistic contact-phenomenon in Gernika Basque. Journal of Language Contact 10(2). 318–352. Romano, Francesco. 2019. Remarks on research of anaphora resolution in situations of language contact: Cross-linguistic influence and the PAS. International Journal of Bilingualism 23(1). 3–20. Romano, Francesco. 2020. Ultimate attainment in heritage language speakers: Syntactic and morphological knowledge of Italian accusative clitics. Applied Psycholinguistics 41(2). 347–380. Romano, Francesco. 2021. L1 versus Dominant Language Transfer Effects in L2 and Heritage Speakers of Italian: A Structural Priming Study. Applied Linguistics 42(5). 945–969. Romano, Francesco. 2023 (this volume). Ultimate attainment of gender in heritage and L2 Italian. In Francesco Romano (ed.), Studies in Italian as a heritage language. Berlin: De Gruyter Mouton. Rothman, Jason, Jorge Alonso & Eloi Puig-Mayenco. 2019. Third Language Acquisition and Linguistic Transfer. Cambridge: Cambridge University Press. Sanchez-Walker, Noelia. 2019. Comprehension of Spanish relative and passive clauses by early bilinguals and second language learners. Unpublished doctoral dissertation. University of Illinois at UrbanaChampaign. Schmidt, Anette. 1985. The fate of ergativity in dying Dyirbal. Language 378–396. Schmitz, Katrin, Laura Di Venanzio & Ana-Lena Scherger. 2016. Null and overt subjects in Italian and Spanish heritage speakers in Germany. Lingua 180. 101–123. Schwartz, Mila. 2008. Exploring the relationship between family language policy and heritage language knowledge among second generation Russian – Jewish immigrants in Israel. Journal of Multilingual and Multicultural Development 29(5). 400–418. Serratrice, Ludovica, Antonella Sorace & Sandra Paoli. 2004. Crosslinguistic influence at the syntax – pragmatics interface: Subjects and objects in English – Italian bilingual and monolingual acquisition. Bilingualism: Language and Cognition 7(3). 183–205. Shain, Cory & Judith Tonhauser. 2010. The synchrony and diachrony of differential object marking in Paraguayan Guaraní. Language Variation and Change 22(3). 321.

32 

 Silvina Montrul

Silva-Corvalán, Carmen. 1994. Language contact and change: Spanish in Los Angeles. Oxford University Press. Smith, Giuditta, Roberta Spelorzi, Antonella Sorace & Maria Garraffa. 2023 (this volume). Similar and divergent: The grammatical competence of adult heritage speakers of Italian compared to adult immigrants on linguistic markers. In Francesco Romano (ed.), Studies in Italian as a heritage language. Berlin: De Gruyter Mouton. Schmitz, Karin & Tim Diaubalick. 2023 (this volume). The expression of (deontic and epistemic) modality in Italian as heritage language in Germany. In Francesco Romano (ed.), Studies in Italian as a heritage language. Berlin: De Gruyter Mouton. Sorace, Antonella. 2011. Pinning down the concept of “interface” in bilingualism. Linguistic Approaches to Bilingualism 1. 1–33. Sorace, Antonella, Ludovica Serratrice, Francesca Filiaci & Maria Baldo. 2009. Discourse conditions on subject pronoun realization: Testing the linguistic intuitions of older bilingual children. Lingua 119. 460–477. Stavans, Anat & Maya Ashkenazi. 2022. Heritage language maintenance and management across three generations: The case of Spanish-speakers in Israel. International Journal of Bilingual Education and Bilingualism 25(3), 963–983. Ticio, Emma. 2015. Differential object marking in Spanish-English early bilinguals. Linguistic Approaches to Bilingualism 5(1). 62–90. Torregrossa, Jacopo & Andrea Listanti. 2023 (this volume). The acquisition of syntactic structures in heritage Italian: Assessing the role of language exposure at critical periods. In Francesco Romano (ed.), Studies in Italian as a Heritage Language. Berlin: De Gruyter Mouton. Tyrrell, Naomi, Pedro Guijarro Fuentes & Claudia Blandon. 2014. Intergenerational family relations and language use: a case study of Spanish-speaking families in Britain. Applied Linguistics Review 5(2). 305–325. Uygun, Serkan & Harald Clahsen. 2021. Morphological processing in heritage speakers: A masked priming study on the Turkish aorist. Bilingualism: Language and Cognition 24(3), 415–426. van Osch, Brechje, Aafke Hulk, Petra Sleeman & Pablo Irizarri van Suchtelen. 2014. Gender agreement in interface contexts in the oral production of heritage speakers of Spanish in the Netherlands. Linguistics in the Netherlands 31(1). 93–106. Von Heusinger, Klaus & Georg Kaiser 2005. In Klaus von Heusinger, Georg Kaiser & Elisabeth Stark (eds.), Proceedings of the Workshop Specificity and the Evolution/Emergence of Nominal Determination Systems in Romance (Arbeitspapier Nr. 119), 33–70. Konstanz: University of Konstanz.

Part I: Experimental studies

Pedro Guijarro Fuentes, Iria Bello Viruega, Estela García Alcaraz, Sergio Viveros Guzmán

Chapter 2  Ultimate attainment in long-immersed heritage Italian immigrants: Syntactic and semantic knowledge of direct object clitics and partitive ne Abstract: Previous studies on heritage language attrition indicate that being exposed to an L2 may have an impact on certain L1 features. Effect incidence levels depend, among other factors, on language competence and grammatical properties. Even if the topic has caught some scholarly attention in recent years, to date, how syntactic complexity may affect morphosyntactic knowledge structures has not been amply investigated. The main purpose of this chapter is to study whether, with unlike to interpretable (i.e., semantic) features, which have been proposed to be a locus of instability, uninterpretable (i.e., syntactic) features are less susceptible to heritage language attrition, and easier to acquire in the L2. To that end, this study investigated knowledge of direct object clitics and partitive ne in Italian in a range of pronominalization direct object constructions in first generation heritage Italian immigrants with considerable exposure to Spanish, that is, those who have been living in Spain for more than five years. Participants were administered, in both Italian and Spanish, a 70-item grammaticality judgement task that contained items with varying degrees of specificity, definiteness, and syntactic complexity. Results were compared to those of L1 (monolingual) Italian and L1 Spanish speakers. In addition, they completed a detailed ethnolinguistic questionnaire in which they were asked to specify their linguistic experience as well as a Spanish and/or Italian overall proficiency test. Results showed that Spanish monolingual speakers’ performance remained in line with the overall expectations and that first-generation heritage speakers of Italian performed like the L1 Spanish group for both semantic and syntactic features, which proved that syntactical representation and semantic knowledge had been fully acquired. Additionally, L2 exposure did not seem to influence the L1, since first-generation heritage participants’ results mirrored those of monolingual Italian speakers when judging the impossibility of object drop. These findings reveal that first-generation heritage Italian immigrants do not show instances of attrition in relation to the linguistic phenomenon

https://doi.org/10.1515/9783110759587-003

36 

 Pedro Guijarro Fuentes et al.

under study, which has a direct impact on the “unaltered” linguistic input that firstgeneration heritage speakers of Italian within this context may receive.1 Keywords: Heritage language attrition, direct object omission, partitive ne, Italian

1 Introduction This chapter offers the results of a study that investigated whether first-generation heritage speakers of Italian (henceforth referred to as heritage speakers within the framework of this study) can attain native-like competence in their knowledge of direct object (DO) omission in Spanish, and whether instances of heritage language attrition effects (loss of structures that had been previously acquired) can be found in the speakers’ L1 regarding this linguistic phenomenon. We tested L2 Spanish, L1 heritage speakers of Italian with more than 5 years of exposure to Spanish in an immersion context in Spain. Within the framework of this research, we understand “heritage speaker” as a bilingual speaker in a minority language immigrant environment and “heritage language” as one of the languages spoken by such an immigrant and previous generations. In the linguistic context investigated herein the community language is Spanish, whereas the minority heritage language is Italian. Therefore, our participants became sequential bilinguals when they started using the community language fluently. The order of language acquisition of the group under study (i.e., Italian first, and Spanish second) and the sociopolitical environment in which our bilingual speakers participated (Spanish majority language vs. Italian minority language) are key features for our inquiry (Montrul 2008 and references thereafter).2 Our population includes a broad range of profiles, with different levels of education and different levels of language dominance and language proficiency in both Italian and Spanish (the minority and the majority

1 We would like to thank the volume editor for his constructive feedback on the initial drafts of the manuscript. Likewise, we appreciate the feedback of the two external experts during the reviewing process. This paper is part of a research project on language attrition and ultimate attainment and ultimate attainment (ref. number: PID2021-122127NB-I00 funded by MCIN/AEI/10.13039/ 501100011033 and by “ERDF A way of making Europe”). 2 We use the term heritage speakers for simplicity, but it actually refers to L1 speakers of Italian as a heritage language. Thus, our participants are in fact preheritage speakers or, in other words, heritage provider speakers. Having just outlined our proposed definition for what we consider to be heritage speakers, we are very much aware that such a definition may seem unorthodox to some.

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 37

languages, respectively). To our knowledge, the association of late L2 acquisition and the attrition of the heritage language has not been previously investigated with the same speakers and with this specific language combination, making this empirical research venue a novel contribution. The study of this linguistic phenomenon has also allowed us to explore L2 acquisition and heritage L1 attrition of interpretable and uninterpretable features. Based on current theorizing in generative research, our working hypotheses are that interpretable features (those contributing essentially to meaning) are more vulnerable to heritage language attrition than non-interpretable features (that is, syntactic features), and that semantic features are easier to acquire than syntactic features (see more in Section 3). Over the years, heritage language attrition has been described as the “disintegration or attrition of the structure of a first language (L1) in contact situations with a second language (L2)” (Seliger and Vago 1991: 3). Thus, it seems natural that bilingual speakers have become the “perfect” candidates in which to investigate the interaction between the L1 and the subsequent language(s). As stated by Vulchanova, Collier, Guijarro-Fuentes and Vulchanov (2022), the study of changes in the language processing and competence of first-generation heritage speakers (i.e., minority language) after being immersed in an L2 context (i.e., majority language) has been gaining momentum in recent decades (see Köpke and Schmid 2004, among others). This line of research has pointed out that immersion in an L2 context can have different effects on L1 competence and, as Vulchanova, Collier, Guijarro-Fuentes and Vulchanov (2022) collect, several scholars have found effects on grammar competence (Håkansson 1995; Gürel 2008) and morphosyntactic features (Montrul 2008, 2016 among others), on speech production and accent (de Leeuw, Schmid and Mennen 2010) as well as on lexical choices (Jarvis 2019). Taking these results into account, speakers of Italian may extend and overgeneralize the use of direct object omission. Thus, it is possible that their heritage L1 grammar is affected while reaching a high attainment in their L2 grammar. Having a comprehensive knowledge of how heritage language attrition works has direct implications for second generation heritage speakers (i.e., heritage speakers for the mainstream heritage literature), since being exposed to a potential attrited heritage language has a direct effect on the development of their heritage language because it constitutes their linguistic input. We chose to examine the L2 acquisition and L1 heritage language attrition of the semantic and syntactical knowledge in heritage speakers of Italian because these areas have been shown to be vulnerable, albeit in a variety of constellations of language combinations and linguistic features such as aspectual contrast and selection of subjunctive mood, to heritage language attrition in adult bilingual speakers (Guijarro-Fuentes and Suárez-Gómez 2021; Montrul 2008; Rothman 2007; Tsimpli, Sorace, Heycock and Filiaci 2004, among others). Previous research

38 

 Pedro Guijarro Fuentes et al.

on heritage languages within the U.S. context (e.g., Bousquette and Putnam 2020; Lohndal, Rothman, Kupisch and Westergaard 2019; Montrul 2016; Polinsky 2011, 2018, a.o.) revealed that L1 and L2 speakers displayed different degrees of grammatical competence by not being exposed to the same quantity and quality of language input. This has led to the discussion within the field whether these competence differences between L1 and L2 speakers could be the result of incomplete/ divergent acquisition (of at least one grammar) (Montrul 2016, a.o.), or language attrition (Schmid and Köpke 2019, a.o.). However, a detailed review of this discussion is beyond the scope of this paper and we refer the reader to Domínguez, Hicks and Slabakova (2009) and references therein. The linguistic property under study in this chapter ˗direct object pronominalization˗ provides evidence of the effect of different grammatical realizations in two languages of the same family. Many languages including Italian and Castilian Spanish allow arguments to be omitted when they are recoverable from the linguistic context due to its semantic feature values. However, while in Castilian Spanish it is possible to omit phrases corresponding to objects which are linked to the semantic nature of the direct object, and not to the syntactic position, as in (1), Italian does not regularly omit object pronouns in these contexts and requires the use of a partitive ne instead, as in (2) (Clements 2006 and references therein; see next section). (1) ¿Vende María peras? Sí, vende. ‘Does María sell pears? Yes, she does.’ (2) Vende Maria delle pere? Sí, ne vende. ‘Does Maria sell pears? Yes, she does.’ The effect of this cross-linguistic variation in argument omission in bilingual language was central in the design of this study. We were interested in investigating to what extent language exposure and practice might influence first-generation heritage speakers’ perception of the partitive ne in Italian (their first language) after being immersed in a Spanish-speaking context for a minimum of 5 years. The remainder of this chapter is set up in the following manner. Section 2 discusses the argument omission in both Italian and Spanish. To that end, we briefly discuss some of the semantic and syntactic similarities and differences associated with the omission of the DO in Spanish compared to Italian. Section 3 presents our main research questions and hypotheses. Section 4 details the design and methodology of the empirical study and it presents the analyses conducted together with the results obtained. Finally, Section 5 discusses our main findings and Section 6 concludes the present study.

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 39

2 Direct object omission and partitive ne To understand how the crosslinguistic element could affect the linguistic competence of adult heritage speakers of Italian, in this section an outline of the linguistic phenomenon under study is proposed. We show that, albeit with different syntactic realization, there is not a crucial semantic difference between Castilian Spanish3 and Italian. The main difference between these two languages concerning direct object pronominalization is that only Castilian Spanish allows their omission if they fulfil some semantic requirements and their referent can be recoverable from the linguistic context. In Castilian Spanish, it is possible to omit arguments corresponding to objects purely based on semantic features (contrast (3a) and (4a) (-definite, -specific) where the DO can be omitted with (3b) and (4b) (+definite +specific) where the DO cannot be omitted, while in Italian the partitive ne is required in equivalent sentences (5a) and (6a) (both (5b) and (6b) (+definite, +specific) require a clitic) (Clements 2006, among others). (3) a. Compra leche todos los días de la semana, pero hoy Buy.3SG milk[-DEF] [-SPEC] every the days of the week but today no Ɵ ha comprado. not have.3SG bought. ‘He/she buys milk every day of the week, but today he/she has not bought any.’ [−definite, −specific] b. Compra la leche todos los días de la semana, pero hoy Buy.3SG the milk[+DEF] [+SPEC] every the days of the week but today no la ha comprado. not it have.3SG bought. ‘He/she buys the milk every day of the week, but today he/she has not bought it.’ [+definite, +specific] (4) a. Compra periódicos todas las semanas, pero Buy.3SG some newspapers [-DEF] [-SPEC] all the weeks but esta semana Ɵ ha olvidado comprar. this week have.3SG forgotten buy ‘He/she buys newspapers every week, but this week he/she has forgotten to buy them.’ [−definite, −specific]

3 Given the variation among different Spanish varieties regarding direct object omission (Masullo 2017, a.o.), for the purpose of this study we only focus on the Spanish spoken in Spain (i.e., Castilian Spanish).

40 

 Pedro Guijarro Fuentes et al.

b. Compra los periódicos todas las semanas, pero Buy.3SG the newspapers [+DEF] [+SPEC] all the weeks but esta semana los ha olvidado comprar. this week them have.3SG forgotten buy ‘He/she buys the newspapers every week, but this week he/she forgot to buy them.’ [+definite, +specific] (5) a. Compra del latte tutti i giorni della settimana, ma Buy.3SG some milk [-DEF] [-SPEC] every the days of the week but oggi non ne ha comprato. today not it have.3SG bought. ‘He/she buys some milk every day of the week, but today he/she has not bought any.’ [−definite, −specific] b. Compra il latte tutti i giorni della settimana, ma Buy.3SG the milk[+DEF] [+SPEC] every the days of the week but oggi no l’ha comprato. today not it have.3SG bought. ‘He/she buys the milk every day of the week, but today he/she has not bought it.’ [+definite, +specific] (6) a. Compra dei giornali ogni settimana, ma Buy.3SG some newspapers [-DEF] [-SPEC] each week but questa settimana se ne é dimenticato comprare. this week it have.3SG forgotten buy ‘He/she buys some newspapers each week, but this week he/she has forgotten.’ [−definite, −specific] b. Compra i giornali ogni settimana, ma Buy.3SG the newspapers [+DEF] [+SPEC] each week but questa settimana se li é dimenticati comprare. this week them have.3SG forgotten ‘He/she buy the newspapers each week, but this week he/she has forgotten them.’ [+definite, +specific] As Rothman and Iverson (2013: 595) state, “phonetically null accusative objects in Spanish show an additional syntactic restriction: As traces of a null variable operator (following Campos 1986 and Huang 1984), which are assumed to move in the syntax, they show evidence of subjacency effects”. Therefore, as Rothman and Iverson (2013) conclude, phonetically dropped objects in standard Castilian

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 41

Spanish are proposed to be banned in strong islands4 (that is, object drop is illicit in strong islands regardless of the definiteness/specificity of the object), as in Determiner Phrase (DP) Islands, Subject Complex Phrase (CP) Islands, and Adjunct Islands, as shown in (7), (8), (9), (10) (11) and (12) respectively.5 (7)

¿Quién encontró dinero? Who found.3SG money [-DEF] [-SPEC]? ‘Who found money?’ [−definite, −specific] No conozco a la persona que ✶Ɵ (lo) encontró. Neg know.1SG DOM the person that ✶Ɵ (it) found.3SG ‘I don’t know the person that found it’.

(8)

¿Trajo Carlos vino a la fiesta? Brought.3SG Carlos wine[-DEF] [-SPEC] to the party? ‘Did Carlos bring wine to the party?’ [−definite, −specific] Que ✶Ɵ (lo) trajo es obvio. That ✶Ɵ (it) brought.3SG is obvious. ‘It’s obvious that he brought it.’

(9)

¿Trajo Carlos vino a la fiesta? Brought.3SG Carlos wine[-DEF] [-SPEC] to the party? ‘Did Carlos bring wine to the party?’ [−definite, −specific] Sí, todos nos emborrachamos porque Yes, all 1ST.PL drunk.1PL because ‘Yes, we all got drunk because he brought it.’

Ɵ (lo) trajo. Ɵ (it) brought.3SG

✶ ✶

(10) Chi ha trovato dei soldi? Who found.3SG money[-DEF] [-SPEC]? ‘Who found money?’ [−definite, −specific] Non conosco la persona che ✶Ɵ (ne) ha trovati. Neg know.1SG DOM the person that ✶Ɵ (it) found.3SG.MASC ‘I don’t know the person that found it’.

4 In generative syntax, an island is a phrase/clause from which you cannot extract a Wh-expression. For instance, in “They told Mary where they wanted to eat lunch” the bolded clause is an island. One cannot extract a Wh-expression out of it that would have lunch as its answer. 5 Please note that the Spanish examples in this section and in Section 4 were taken from Bruhn de Garavito and Guijarro-Fuentes (2001).

42 

 Pedro Guijarro Fuentes et al.

(11) Ha portato Carlo del vino a la festa? Brought.3SG Carlos some wine [-DEF] [-SPEC] to the party? ‘Did Carlos bring wine to the party?’ [−definite, −specific] Che ✶Ɵ (ne) ha portato è quello che ho sentito. That ✶Ɵ (it) brought.3SG is obvious. ‘It’s obvious that he brought it.’ (12) Ha portato Carlo del vino a la festa? Brought.3SG Carlo wine[-DEF] [-SPEC] to the party? ‘Did Carlos bring wine to the party?’ [−definite, −specific] ✶ Si, ci siamo tutti ubriacati perché Ɵ (ne) ha portato Yes, 1ST.PL all drunk.1PL.MASC because ✶Ɵ (it) brought.3SG ‘Yes, we all got drunk because he brought it.’ In turn, Italian does not allow object drop, and it implies to exhibit island sensitivity, similarly to what has been suggested for Castilian Spanish (see examples 10, 11, and 12). Given the similarities and differences between Italian and Spanish (see Table 1), this Romance pair provides an interesting opportunity for issues of L2 ultimate attainment and heritage L1 language attrition. Table 1: Summary of property asymmetries for object drop in Italian and Castilian Spanish. Italian

Castilian Spanish

Object drop [+ definite, +specific, interpretable feature]

No

No

Object drop [-definite, –specific, interpretable feature]

Does not apply. Use of partitive ne

Yes

Subjacency effects on object drop [uninterpretable [+Top] feature]

Yes

Yes

3 Current study: Research questions and hypotheses Previous research on near-nativeness in adult L2 acquisition is scarce. However, it shows that although adult sequential bilinguals may display native-like proficiency, their grammatical mental representation seems to differ from that of monolingual speakers (Schmid and Yilmaz, 2018, a.o.). For example, as Tsimpli, Sorace,

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 43

Heycock and Filiaci (2004) points out, it has been shown that specific syntactic constraints which seem to be categorical for native speakers appear to be optional for near-native speakers (Robertson and Sorace 1999, a.o.). In addition, there has been little research (e.g., Tsimpli 2014 and references therein) on possible effects on the advanced L2 near-native grammar in heritage language groups; basically because of the assumption that an L2 does not directly affect the L1 grammar (e.g., White 2003; Sorace 2005). To the best of our knowledge, the reciprocal influence of the two grammar systems of near-native speakers of two languages of the same family (Romance languages in this study) has received so far little scholarly attention. Despite the fact that many scholars have devoted their research to investigating language attrition (e.g., Bardovi-Harlig and Stringer 2010), as Sorace (2005: 67) defends, this research is only partly relevant here, since most of it, especially those first studies, has been concerned “with migrant communities, which are usually characterized by diminished use of the L1, separation from the L1 speaking community, low degree of acculturation, and a low-level of L2 attainment, at least in the first generation”. More relevant for our purposes are a limited case studies research papers, which, according to Sorace (2005: 67) have made possible to propose the following two descriptive generalizations (see Sorace’s paper for a full review): (a) The selectiveness of attrition process due to different degrees of vulnerability in distinct grammatical aspects. (b) The loss of more complex rules in favor of those less complex. However, empirical research devoted to the study of language attrition within generative linguistic approaches is still limited. Tsimpli (2014) argued that while interpretable features seemed to be vulnerable to language attrition, uninterpretable features would be more difficult to “deteriorate”. In support of this hypothesis, Tsimpli, Sorace, Heycock and Filiaci (2004) showed that the uninterpretable feature that is responsible for subject realization in L1 Italian and Greek is immune to the influence of L2 English. Nevertheless, neither of these studies have considered adult sequential bilingual speakers of the same family language. The comparison of the possible heritage language attrition of a Romance language (i.e., Italian) due to the L2 acquisition of another Romance language (i.e., Castilian Spanish) is a relatively unexplored field in heritage language acquisition research. Thus, this novel research aims to contribute to this line of research by exploring the grammatical competence in both the Italian and Spanish of heritage speakers of Italian living in Spain. To do so, it is essential to consider the potential influence of the L1, which could show signs of attrition, in the grammatical knowledge of sequential bilingual

44 

 Pedro Guijarro Fuentes et al.

speakers who have been exposed to their L2 for a lengthy period of time. Within this perspective, a study of sequential heritage bilingual speakers may shed light on contact-induced change among L1 Romance speakers. Against this backdrop, in this study the following research questions are considered: RQ1: How similar is the language competence in relation to DO realization/omission in long-immersed heritage immigrant and Spanish and Italian monolingual speakers, particularly when a structure is not present in one of the languages and the semantic/syntactic information is partly similar? RQ2: To what degree does the L2 (Spanish) affect the L1 (Italian) of long-immersed heritage immigrant grammars? To answer both research questions, our participant sample involved adult heritage bilingual speakers with age of L2 onset after puberty. Our main prediction was that high attainment of direct object omission in their L2 would affect the partitive ne in first generation heritage speakers of Italian. Based on some of the research outlined previously, the central theoretical hypothesis guiding this paper is that both differential L2 ultimate attainment and heritage language attrition depend on the nature of the features involved. Being more specific, we hypothesized that interpretable (semantic) features, different to non-interpretable features (narrow syntax) (see Chomsky 1995, 2001), are more susceptible to heritage language attrition. This is exactly the inverse of what is predicted by the Interpretability Hypothesis (Hawkins and Hattori 2006; Tsimpli and Dimitrakopoulou 2007, a.o.), as in this model, deficits are associated with uninterpretable features, not interpretable ones. Thus, we predicted a kind of mirror image scenario, where properties not affected by attrition in the L1 are those which are more likely to be a challenge in the L2. Our hypothesis also runs counter to the Interface Hypothesis (Sorace 2011, 2012), which assumes that the same areas that are more susceptible in L2 acquisition are also vulnerable in domains such as heritage language attrition. Thus, to confirm our main hypothesis, we aim to obtain reliable data concerning the acceptability of sentences that are only consistent with one or the other grammar. This way, we will be in a better position to understand whether heritage speakers are really accessing two fully grammatical options. In the next section, we describe our methodology together with our analyses and main results.

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 45

4 Method 4.1 Participants This study involved three groups of participants: (i) a group of 10 heritage speakers of Italian that were also very advanced L2 Spanish speakers and that had been living in Spain for more than five years. The L2 exposure of the heritage language group began in their late puberty or adulthood (> 15 years old) and all of them obtained a minimum score of 43 (out of 50) in the Spanish proficiency test administered; (ii) a control group of 12 Spanish native speakers residing in Spanish monolingual autonomous communities6 (Spanish monolingual group), and (iii) another control group of 10 Italian native speakers living in Italian monolingual regions (Italian monolingual group). The group of heritage speakers of Italian was tested in both Italian and Spanish, whereas the native speakers (NSs) completed the task only in their L1. In the heritage Italian group there were seven males and three females with an age range between 27–45 years old (Mean: 36.10; SD: 5.84), in the Spanish monolingual group there were six males and six females with an age range between 21–52 years old (Mean: 36.58; SD: 9.22) and in the Italian monolingual group there were five males and five females with an age range between 30–71 years old (Mean: 43.00; SD: 14.19). All participants had at least completed secondary education. Table 2 summarizes our participants’ pool. Table 2: Participants’ pool. Heritage group

Spanish monolinguals

Italian monolinguals

Number of participants

10

12

10

Languages spoken

L1 Italian, L2 Castilian Spanish

L1 Spanish

L1 Italian

Residence

Spain

Spain

Italy

Age (mean)

36.10

36.58

43.00

6 Spain is divided administratively in 17 autonomous communities. While Spanish is the official language in the whole country, some autonomous communities also have a co-official language (e.g., Catalan, Basque, or Galician) and are considered bilingual territories.

46 

 Pedro Guijarro Fuentes et al.

4.2 Stimuli and procedure All participants were administered a five-point Likert scale grammaticality judgment task adapted from Bruhn de Garavito and Guijarro-Fuentes (2001). In this task, participants were presented with a question followed by a potential answer to the question and they were asked to judge how natural they perceived the answer given in a scale from 1 (totally unnatural) to 5 (totally natural). Ten conditions were included. Condition 2, which contained an ungrammatical utterance with a gender-mismatch, served as a control variable. The first four conditions tested four different variables in simple structures, namely definiteness and specificity of the antecedent, presence/absence of object clitic/pronoun ne, as well as grammaticality, as shown in (11) to (14). More specifically, Condition 1 (example 13) contained grammatical sentences in which the object (+definite/+specific) was not dropped. In  Condition 2 (example 14), the same sentences were made ungrammatical by altering the expected gender agreement between the pronoun and its referent. Condition 3 (example 15) was made up of grammatical sentences in which the object (-definite/-specific) was dropped in Spanish but kept in Italian (ne), whereas in Condition 4 (example 16) the drop of the object (+definite/+specific) made the sentence ungrammatical in both languages. (13) Simple clause, definite and specific antecedent, overt object clitic, grammatical sentence a. ¿Trajo Pedro los regalos para la fiesta? 12345 ‘Did Pedro bring the gifts for the party?’ Sí, los trajo. ‘Yes, he brought them.’ b. Ha portato Piero i regali per la festa? 12345 ‘Did Piero bring the gifts for the party?’ Sì, li ha portati. ‘Yes, he brought them.’ (14) Simple clause, definite and antecedent, overt object clitic, ungrammatical sentence (gender mismatch) a. ¿Trajo Pedro los regalos para la fiesta? 12345 ‘Did Pedro bring the gifts (masc.) for the party?’ Sí, ✶las (fem) trajo. ‘Yes, he brought them.’ b. Ha portato Piero i regali per la festa? 12345 ‘Did/Piero bring the gifts (masc.) for the party?’ Sì, ✶le (fem) ha portate. ‘Yes, he brought them.’

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 47

(15) Simple clause, indefinite and non-specific antecedent, object clitic dropped in Spanish/overt pronoun ne kept in Italian, grammatical sentence. a. ¿Trajo Pedro regalos para la fiesta? 12345 ‘Did Pedro bring gifts to the party?’ Sí, trajo. ‘Yes, he brought some.’ b. Ha portato Piero dei regali per la festa? 12345 ‘Did Piero bring gifts to the party?’ Sì, ne ha portati. ‘Yes, he brought some.’ (16) Simple clause, definite and antecedent, object dropped, ungrammatical sentence a. ¿Trajo Pedro los regalos para la fiesta? 12345 ‘Did Pedro bring gifts to the party?’ Sí, ✶Ɵ trajo. ‘Yes, he brought some.’ b. Ha portato Piero i regali per la festa? 12345 ‘Did Piero bring gifts to the party?’ Sì, ✶Ɵ ha portati. ‘Yes, he brought some.’ The remaining six conditions tested participants’ acceptability of direct object clitics and partitive  ne  in complex  constructions: the structure was included in strong islands functioning as DPs (Conditions 5 and 6, examples 17 and 18, respectively), Subjects (Conditions 7 and 8, examples 19 and 20, respectively), and Adjuncts (Conditions 9 and 10, examples 21 and 22, respectively). In these sentences, the overt clitic option was the only grammatical choice. As a result, Conditions 6, 8, and 10 presented grammatical options where the pronoun was present regardless of the definiteness and specificity of the object. Conversely, in Conditions 5, 7, and 9 ungrammaticality was achieved by dropping the pronoun. (17) Determiner Phrase (DP) Island, definite and antecedent, object dropped, ungrammatical sentence a. ¿Quién trajo los regalos para la fiesta? 12345 ‘Who brought the gifts for the party?’ No conozco al muchacho que ✶Ɵ trajo. ‘I don’t know the boy who brought.’

48 

 Pedro Guijarro Fuentes et al.

b. Chi ha portato i regali per la festa? ‘Who brought the gifts for the party?’ Non conosco il ragazzo che ✶Ɵ ha portati ‘I don’t know the boy who brought.’

12345

(18) Determiner Phrase (DP) Island, definite and antecedent, overt object clitic, grammatical sentence a. ¿Quién trajo los regalos para la fiesta? 12345 ‘Who brought the gifts to the party?’ No conozco a la persona que los trajo. ‘I don’t know the person who brought them.’ b. Chi ha portato i regali per la festa? 12345 ‘Who brought the gifts to the party?’ Non conosco la persona che li ha portati. ‘I don’t know the person who brought them.’ (19) Subject CP Island, indefinite and non-specific antecedent, object dropped, ungrammatical sentence a. ¿Trajo Pedro regalos para la fiesta? 12345 ‘Did Pedro bring gifts to the party?’ Que ✶Ɵ trajo es lo que oí. ‘That he brought is what I heard.’ b. Ha portato Piero dei regali per la festa? 12345 ‘Did Piero bring gifts to the party?’ Che ✶Ɵ ha portati è quello che ho sentito. ‘That he brought is what I heard.’ (20) Subject CP Island, definite and non-specific antecedent, overt object clitic, grammatical sentence a. ¿Trajo Pedro regalos para la fiesta? 12345 ‘Did Pedro bring gifts to the party?’ Es obvio que los trajo. ‘It is obvious that he brought some.’ b. Ha portato Piero dei regali per la festa? 12345 ‘Did Piero bring gifts to the party?’ È ovvio che ne ha portati ‘It is obvious that he brought some.’

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 49

(21) Adjunct Island, indefinite and non-specific antecedent, object dropped, ungrammatical sentence a. ¿Trajo Pedro regalos para la fiesta? 12345 ‘Did Pedro bring gifts to the party?’ Sí, todos nos alegramos porque ✶Ɵ trajo. ‘Yes, we were all glad because he brought.’ b. Ha portato Piero dei regali per la festa? 12345 ‘Did Piero bring gifts to the party?’ Sì, eravamo tutti contenti perché ✶Ɵ ha portati. ‘Yes, we were all glad because he brought.’ (22) Adjunct Island, definite and non-specific antecedent, overt object clitic, grammatical sentence a. ¿Trajo Pedro regalos para la fiesta? 12345 ‘Did Pedro bring gifts to the party?’ Sí, todos nos alegramos porque trajo muchos. ‘Yes, we were all pleased because he brought so many.’ b. Ha portato Piero dei regali per la festa? 12345 ‘Did Piero bring gifts for the party?’ Sì, eravamo tutti contenti perché ne ha portati. ‘Yes, we were all glad because he brought so many.’ Every condition was replicated seven times. This made up a total of 70 experimental items that were presented in a randomized order. All experimental items shared a set of features: they had a singular, human subject and an action verb conjugated in the past. Masculine and feminine objects were counterbalanced, and the same seven verbs were used in each condition. Besides, the DO was always an object in the plural, and it always contained a short postmodifying PP. Word frequency and register were controlled and polysemy was avoided, as suggested by Clifton, Staub and Rayner (2007). Two equivalent versions of the task (one in Spanish and one in Italian) were created. The experimental sentences were presented to participants as part of a survey created in Google Forms. All participants were contacted by e-mail. They received a link to the survey in their L1, which they completed in their own electronic devices without time limitations. Bilingual speakers completed the Italian version first and, after a week, they received the Spanish version of the survey. All participants took part in the experiment voluntarily and signed a consent form before commencing the experiment.7 7 The consent form complied with the regulations of the University of the Balearic Islands (UIB) Research Ethics Committee of May 13, 2021 (File number 206CER21), created through the Regulatory Agreement of June 13, 2014 (FOU 403).

50 

 Pedro Guijarro Fuentes et al.

The survey was structured in several sections. After signing the consent form, participants answered an ethnolinguistic questionnaire specifying their gender, age, place of residence, etcetera. After that, participants were asked to complete an adaptation of the Bilingual Language Profile (Birdsong, Gertken and Amengual, 2012) to assess language dominance in the pair of languages under study (Spanish and Italian). Then a proficiency test was included to assess participants’ linguistic knowledge. In the Spanish version, participants completed two multiple-choice tasks:8 in the first one, a set of 30 sentences containing a gap together with 4 possible answers was included. In the second activity, participants read a text containing 20 gaps and were given three possible answers per item. The Italian version (Kupisch, Barton, Bianchi and Stangen 2012) contained a text with 45 gaps that participants needed to fill with a single word of their choice. Once the proficiency test was completed, participants could read the instructions to the grammaticality judgment task and eventually complete it. Most participants completed the survey in 30–45 minutes. A summary of the different conditions tested is provided in Table 3. Table 3: Conditions tested. Grammaticality Antecedent Pronoun Simple Determiner Subject Adjunct Definiteness/ use clause Phrase (DP) CP Island Island Specificity Island Condition 1 Condition 2 Condition 3 Condition 4 Condition 5 Condition 6 Condition 7 Condition 8 Condition 9 Condition 10

+ − + − − + − + − +

+ + − + + + − − − −

+ + –SP/+IT − − + − + − +

+ + + + − − − − − −

− − − − + + − − − −

− − − − − − + + − −

− − − − − − − − + +

8 As detailed in Cuza and López Otero (2016: 471), this proficiency test, which was developed by White and colleagues (Duffield and White 1999), consists of two sections: a vocabulary task from the MLA Foreign Language Test and a cloze test adapted from the Diploma de Español como Lengua Extranjera (DELE) test.

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 51

4.3 Analyses and results We recruited a total of 38 participants (12 heritage speakers of Italian, 12 Spanish monolinguals and 14 Italian monolinguals). However, two participants were excluded from the heritage group because they did not meet the experimental group selection criteria (heritage speakers of Italian who were exposed to L2 Spanish at late puberty or during adulthood, had resided a minimum of five years in Spain and had an advanced level of Spanish acknowledged by proficiency test scores). Two additional Italian monolingual participants were not included in the final monolingual samples since they were living in a country other than Spain or Italy. To ensure data reliability, the participants’ mean for Condition 2 was calculated9 and two additional Italian monolingual participants were excluded from the final pool. Their mean for Condition 2 was higher than 3.5, which may indicate high lack of proficiency and/or lack of focus because they were not able to detect the presented ungrammaticality. The results for Condition 2 will be presented but they will not be analyzed any further, as they were used in this study as a control variable. As already specified in Section 4.1, the final sample consisted of 10 heritage speakers of Italian who completed the task in both Italian and Spanish, 12 Spanish monolinguals and 10 Italian monolinguals. Regarding the analyses and coding of the data for both linguistic tasks, judgments on the Likert Scale were translated into means per participant per condition. Both descriptive and inferential statistics analysing the monolingual and heritage speakers’ performance on the experimental task are presented below. Both paired samples t-tests and independent samples t-tests were run to check within and between, respectively, subject differences. Statistical analyses were run in JASP (version 0.14.1) (JASP Team 2010) and the alpha level was set at .05. Results are organized as follows: first, Spanish monolingual data is presented followed by Italian monolingual data. Sentences targeting interpretable features (Conditions 1–4) and uninterpretable features (Conditions 5–10) are presented separately. Next, results from heritage speakers of Italian are compared to those of Italian and Spanish monolingual groups. It is important to highlight that for the purpose of this study a mean superior to 3 is interpreted as a preference and a mean under this cut-off as a rejection.

9 Condition 2 is a control variable in which a gender mismatch between the referent and the clitic is included.

52 

 Pedro Guijarro Fuentes et al.

Spanish monolingual data: Interpretable features As shown in Table 4, the results obtained for Spanish monolingual speakers are in line with what has been previously described for Castilian Spanish. Spanish monolinguals judged as inacceptable those sentences including a null object referring to a specific and definite referent (Condition 4), while they judged as acceptable those sentences including a null object referring to a no specific and no definite referent (Condition 3). The highest rate of acceptance is for sentences including an overt object pointing to a definite and specific referent (Condition 1). Table 4: Interpretable features: Spanish monolingual speakers. Condition 1 Participants Mean Std. Deviation

12 4.988 0.041

Condition 2

Condition 3

12 1.262 0.658

12 3.488 1.194

✶

Condition 4

✶

12 2.250 0.912

NB: Condition 1: the object (+definite/+specific) is not dropped = grammatical (gender match); Condition 2: the object (+definite/+specific) is not dropped = ungrammatical (gender mismatch); Condition 3: the object (– definite/– specific [and generic] is dropped = grammatical; Condition 4: the object (+definite/+specific [and generic]) is dropped = ungrammatical. Ungrammaticality is marked with✶.

Spanish monolinguals: Uninterpretable features In Castilian Spanish, to drop the object in complex islands is considered ungrammatical regardless of the definiteness and specificity of the referent. To check whether our Spanish monolingual speakers followed this pattern, we presented them with sentences containing either null or overt objects in complex DP, Subject, and Adjunct islands. The results presented in Table 5 show that Spanish monolinguals accepted more those sentences containing an overt object rather than a null object in DP islands (t (11) = −8.748, p < .001, d = −2.52) and Adjunct islands (t (11) = −4.996, p < .001, d = −1.42). The case of Subject islands seems to be similar, since when comparing the participants’ mean for condition 7 and 8, a paired samples t-test revealed a marginal significant difference with a medium size effect (t (11) = −2.079, p = .062, d = 0.60). Thus, our Spanish monolingual speakers seem to align with the pattern of Castilian Spanish.

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 53

Table 5: Uninterpretable features: Spanish monolingual speakers. Condition 5 Condition 6

Participants Mean Std. Deviation

✶

12 1.690 0.683

12 4.452 0.830

Condition 7 Condition 8

✶

12 2.679 1.003

Condition 9 Condition 10

✶

12 3.369 1.020

12 2.667 0.883

12 4.238 1.027

NB: Condition 5: Complex DP with null object = ungrammatical (referent: +definite +specific); Condition 6: Complex DP with overt object = grammatical (referent: +definite +specific); Condition 7: Sentential Subject with null object = ungrammatical (referent: –definite –specific); Condition 8: Sentential Subject with overt object = grammatical (referent: –definite –specific); Condition 9: Adjunct with null object = ungrammatical (referent: –definite –specific); Condition 10: Adjunct with overt object = grammatical (referent: –definite –specific). Ungrammaticality is marked with✶.

Italian monolingual data: Interpretable features Table 6: Interpretable features: Italian monolingual speakers. Condition 1 Participants Mean Std. Deviation

10 4.229 0.684

Condition 2

Condition 3

10 1.671 0.738

10 3.029 1.071

✶

Condition 4

✶

10 1.486 0.754

NB: Condition 1: the object (+definite/+specific) is not dropped = grammatical (gender match); Condition 2: the object (+definite/+specific) is not dropped = ungrammatical (gender mismatch); Condition 3: the object (– definite/– specific [and generic] is recovered with the clitic ne = grammatical; Condition 4: the object (+definite/+specific [and generic]) is dropped= ungrammatical. Ungrammaticality is marked with✶.

According to the data contained in Tables 4 and 6, both Spanish and Italian groups showed similar acceptance rates for the same experimental conditions. To confirm this, independent samples t-test were run for each experimental condition comparing the mean in the Spanish and Italian groups. Significant differences were found for Condition 1 (t (20) = −3.857, p < .001, d = −1.65) and Condition 4 (t (20) = −2.114, p = .047, d = −.91). However, it is important to note that these results do not show different patterns of behaviour but differences in terms of “strength of preferences”: the acceptance rate for Condition 1 is stronger in Spanish albeit in Italian it also surpasses 4 and the rejection rate for Condition 4 is higher in Italian although in Spanish it barely surpasses 2.

54 

 Pedro Guijarro Fuentes et al.

Italian monolingual data: Uninterpretable features Table 7: Uninterpretable features: Italian monolingual speakers. Condition 5 Condition 6

Participants Mean Std. Deviation

✶

10 2.029 0.973

10 4.943 0.181

Condition 7 Condition 8

✶

10 1.571 0.891

10 3.043 1.163

Condition 9 Condition 10

✶

10 1.643 0.671

10 3.129 1.400

NB: Condition 5: Complex DP with null object = ungrammatical (referent: +definite +specific); Condition 6: Complex DP with overt object = grammatical (referent: +definite +specific); Condition 7: Sentential Subject with null object = ungrammatical (referent: –definite –specific); Condition 8: Sentential Subject with overt object (ne) = grammatical (referent: –definite –specific); Condition 9: Adjunct with null object = ungrammatical (referent: –definite –specific); Condition 10: Adjunct with overt object (ne)= grammatical (referent: –definite –specific). Ungrammaticality is marked with✶.

If Tables 5 and 7 are compared, it can be noted that overall acceptability patterns for complex islands in both languages are similar. Grammaticality seems to be a decisive issue, as in both Spanish and Italian groups, Conditions 6, 8, and 10 (grammatical sentences) were deemed more acceptable than Conditions 5, 7, and 9 (ungrammatical sentences). Independent samples t-tests comparing participants’ mean per condition in both Spanish and Italian showed significant differences for Condition 7 (t (20) = −2.710, p = .013, d = −1.160), Condition 9 (t (20) = −3.008, p =.007, d = −1.288), and Condition 10 (t (20) = −2.142, p = .045, d = −.917). Again, differences do not suggest different patterns of behaviour between languages but different degrees of strength. Overall, the reaction to object drop seems to be similar in both languages for interpretable and uninterpretable features. Now, the comparisons between the heritage group and Spanish monolinguals, on the one hand, and between the heritage group and Italian monolinguals, on the other hand, are presented.

Heritage speakers of Italian vs. Italian monolingual speakers: Interpretable and Uninterpretable features As presented in Tables 8 and 9, both Italian monolinguals and heritage speakers of Italian showed rather similar rates of acceptance for sentences evaluating both interpretable and uninterpretable features. To check whether inferential statistics confirmed that both groups performed similarly when judging object drop in Italian, independent samples t-tests were run for each experimental condition. The results show that both groups perform similarly for all experimental conditions (p ≥ .096). However, although results for Condition 3 do not show a significant effect, a

 55

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

Table 8: Interpretable features: Heritage vs. monolingual speakers of Italian. Condition 1 Participants Mean Std. Deviation

Condition 2

✶

Condition 3

Condition 4

✶

HER_IT

MON_IT

HER_IT

MON_IT

HER_IT

MON_IT

HER_IT

MON_IT

10 4.443 0.604

10 4.229 0.684

10 1.400 0.605

10 1.671 0.738

10 2.129 1.219

10 3.029 1.071

10 1.400 0.746

10 1.486 0.754

NB: Condition 1: the object (+definite/+specific) is not dropped = grammatical (gender match); Condition 2: the object (+definite/+specific) is not dropped = ungrammatical (gender mismatch); Condition 3: the object (– definite/– specific [and generic] is recovered with the clitic ne = grammatical; Condition 4: the object (+definite/+specific [and generic]) is dropped= ungrammatical. Ungrammaticality is marked with✶. HER-IT: Heritage group in Italian; MON-IT: Italian monolingual group.

marginal significant effect with medium effect size (t (18) = −1.754, p =.096, d = −.785), was found, which could be pointing to a lower acceptancy rate for bilingual speakers when using the clitic ne to recover a generic object. It is premature to draw such conclusion considering that the statistical analysis did not reveal a clear difference between both groups, but this is something we need to study further with a larger sample in future studies. At this time and, considering the results obtained, heritage speakers do not seem to show signs of attrition and the use of ne is well retained in their heritage grammar despite being intensively exposed to a L2 that differs in this respect.

Heritage speakers of Italian vs. Spanish monolingual speakers: Interpretable and Uninterpretable features As it was the case for Tables 8 and 9, Tables 10–11 show that both Spanish monolinguals and heritage speakers of Italian perform quite similarly when evaluating the acceptance of object drop in sentences targeting both interpretable and uninterpretable features. To allow statistical confirmation, independent samples t-tests were run for each experimental condition. Following the pattern found in Italian, results show that both groups (HER-IT and MON-SP) performed similarly in Spanish for all experimental conditions (p ≥ .106). These results show that the Spanish-Italian bilinguals in our sample have attained native like performance in relation to the object drop linguistic phenomenon in their second language. In the next section, a discussion of the interpretation of these results in light of our research questions and of their implications for current theorizing generative accounts on ultimate attainment and heritage language attrition is presented.

10 1.429 0.828

HER_IT 10 2.029 0.973

MON_IT 10 4.829 0.231

HER_IT 10 4.943 0.181

MON_IT

Condition 6 10 1.429 0.343

HER_IT 10 1.571 0.891

MON_IT

Condition 7

✶

10 2.586 1.167

HER_IT 10 3.043 1.163

MON_IT

Condition 8 10 1.300 0.378

HER_IT 10 1.643 0.671

MON_IT

Condition 9

✶

10 2.886 1.164

HER_IT

10 3.129 1.400

MON_IT

Condition 10

NB: Condition 5: Complex DP with null object = ungrammatical (referent: +definite +specific); Condition 6: Complex DP with overt object = grammatical (referent: +definite +specific); Condition 7: Sentential Subject with null object = ungrammatical (referent: –definite –specific); Condition 8: Sentential Subject with overt object (ne)= grammatical (referent: –definite –specific); Condition 9: Adjunct with null object = ungrammatical (referent: -definite -specific); C10: Adjunct with overt object (ne)= grammatical (referent: -definite -specific). Ungrammaticality is marked with✶. HER-IT: Heritage group in Italian; MON-IT: Italian monolingual group.

Participants Mean Std. Deviation

 

✶

Condition 5

Table 9: Uninterpretable features: Heritage vs. monolingual speakers of Italian.

56   Pedro Guijarro Fuentes et al.

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 57

Table 10: Interpretable features: Heritage vs. Spanish monolingual speakers. Condition 1 Participants Mean Std. Deviation

Condition 2

✶

Condition 3

Condition 4

✶

HER-IT

MON_SP

HER-IT

MON_SP

HER-IT

MON_SP

HER-IT

MON_SP

10 4.986 0.045

12 4.988 0.041

10 1.529 0.628

12 1.262 0.658

10 2.629 1.175

12 3.488 1.194

10 2.200 1.147

12 2.250 0.912

NB: Condition 1: the object (+definite/+specific) is not dropped = grammatical (gender match); Condition 2: the object (+definite/+specific) is not dropped = ungrammatical (gender mismatch); Condition 3: the object (– definite/– specific [and generic] is dropped= grammatical; Condition 4: the object (+definite/+specific [and generic]) is dropped= ungrammatical. Ungrammaticality is marked with✶. HER-IT: Bilingual group in Spanish; MON-SP: Spanish monolingual group.

5 General discussion The Grammaticality Judgement tasks in both in Spanish and Italian were designed to assess Spanish monolingual and heritage speakers of Italian’s preference regarding the possibility of object drop in Spanish based on both semantic and syntactic constrains. Additionally, we also intended to assess Italian monolingual and heritage speakers of Italian’s intuitions on the acceptability of the partitive ne pronoun in Italian. Our main prediction was that long-immersed heritage bilingual speakers would converge with Spanish native speakers. In other words, we did not expect cross-linguistic effects even if bilinguals must have acquired a new syntactic option for licensing drop object in Spanish and they still needed to respect partitive ne use in all syntactic structures that required it in Italian. Statistical results pointed to similar performances between Spanish and Italian and between monolingual and sequential bilingual speakers. Both Spanish and Italian monolingual as well as heritage speakers of Italian performed comparably, and the three groups met the expectations laid out by theoretical and descriptive studies. This leads us to the conclusion that heritage speakers of Italian with a long immersion in Spain (> 5 years) can attain native-like performance in their L2 while keeping their L1 heritage competence intact. Regarding object drop and partitive ne, the bilingual group deviated more often from previous accounts which claimed that, while the interpretable (i.e., semantic) features are the locus of instability, uninterpretable (i.e., syntactic) features are less susceptible to heritage language attrition, and easier acquired in the L2. As for both monolingual control groups, behaviour was systematic. This is true for the Spanish monolinguals given that the results of the Italian monolingual group turned out to be inconclusive with a certain heterogeneity regarding condition 3 (more on this below). We could conclude that mono-

10 1.729 0.801

12 1.690 0.683

10 4.071 1.119

HER-IT 12 4.452 0.830

MON_SP

Condition 6 10 2.129 1.237

12 2.679 1.003

MON_SP

Condition 7

HER-IT

✶

10 3.314 1.409

HER-IT 12 3.369 1.020

MON_SP

Condition 8 10 2.043 0.841

12 2.667 0.883

MON_SP

Condition 9

HER-IT

✶

10 4.529 0.667

12 4.238 1.027

MON_SP

Condition 10 HER-IT

NB: Condition 5: Complex DP with null object = ungrammatical (referent: +definite +specific); Condition 6: Complex DP with overt object = grammatical (referent: +definite +specific); Condition 7: Sentential Subject with null object = ungrammatical (referent: –definite –specific); Condition 8: Sentential Subject with overt object = grammatical (referent: -definite -specific); Condition 9: Adjunct with null object = ungrammatical (referent: -definite -specific); Condition 10: Adjunct with overt object = grammatical (referent: -definite -specific). Ungrammaticality is marked with✶. HER-IT: Bilingual group in Spanish; MON-SP: Spanish monolingual group.

Participants Mean Std. Deviation

MON_SP

Condition 5

HER-IT

✶

Table 11: Uninterpretable features: Heritage vs. Spanish monolingual speakers.

58   Pedro Guijarro Fuentes et al.

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 59

lingual and sequential bilingual speakers have a tendency to meet expectations in both core and less core cases. The relevance of our results for ultimate attainment and heritage language attrition needs to be discussed very circumspectly given some of the limitations of our study; we next, nevertheless, seek answers to our research questions. RQ1: How similar is the language competence in relation to DO realization/omission in long-immersed heritage immigrant and Spanish and Italian monolingual speakers, particularly when a structure is not present in one of the languages and the semantic/syntactic information is partly similar?

There are not significant differences between the L1 Spanish and Italian groups. Generally speaking, the behaviour detected for both Spanish and Italian monolinguals agreed with what has been previously described by theoretical studies (see Campos 1986 and Clements 2006 for Spanish). Similarly to Rothman and Iverson (2013), our data reveals that the (im)possibility to omit the object in Spanish and the partitive ne in Italian are subject to both semantic and syntactic restrictions. As a result, according to previous descriptive data, monolingual speakers should not accept as feasible sentences including the omission of the object or of the partitive ne in contexts -definite/-specific in simple sentences or in syntactic islands. Overall, our results replicate findings from the previous empirical literature with different populations using similar linguistic tasks (Bruhn de Garavito and Guijarro-Fuentes 2001; Rothman and Iverson 2013, a.o). It is interesting to note that the Italian monolingual group showed a certain degree of variability in those conditions including the partitive ne in grammatical contexts. These findings were striking, since partitive articles as well as the partitive clitic ne are part of the standard Italian grammatical system. A possible explanation could be found in Gerards and Stark (2021)’s remark that the use of partitive articles is more regular in the northern varieties. Diatopic variation could indeed account for the variability of our results, and it would be, somehow, natural to expect certain irregularity when the use of ne is a stake, since one of its functions is to recover a previous partitive article. In a similar vein, Mariotti and Nissim (2014: 249)’s corpus study showed that although “the clitic ne is traditionally described as a partitive particle [. . .] the partitive feature is not the dominant”. Their results showed that only 14.2% of the 500 instances of ne found in their corpus were partitive uses. This finding can serve as an indication of low Italian speakers’ exposure to this feature, and it could add up to the aforementioned evidence of diatopic variation to be used as an argument to explain the variability of our results regarding the use of ne. Future research should focus on disentangling whether the optionality observed in our monolingual Italian group is maintained with a larger sample including balanced groups of Italian speakers from both northern and central-southern varieties.

60 

 Pedro Guijarro Fuentes et al.

RQ2: To what degree does the L2 (Spanish) affect the L1 (Italian) of long-immersed heritage immigrant grammars?

Contra to our expectations, the results indicated a similarity between heritage and monolingual speakers of Italian. No signs of heritage language attrition were found. Thus, L1 heritage speakers’ competence remained intact. The extended exposure to a language lacking an equivalent form to the partitive ne like Spanish proved to have no effect in our sample. It has been proposed that both frequency and domains of usage could have an impact on language attrition (Schmid and Yılmaz 2018). Nevertheless, as stated in Vulchanova et al. (2022) following Schmid (2019), there is an interactive complex scenario among external and other predictor variables (i.e., frequency and domains of usage) that makes feasible that a possible explanation might be found beyond external factors. Notwithstanding the “low” acceptancy rate (around 3) that monolinguals showed for Condition 3, the rate was even lower, although not statistically different, for bilingual speakers. Future research should confirm this trend with a larger sample. In the meantime, taking into account that our participants were monolingual speakers of standard Italian, our results seem to be, somehow, similar to Perpiñan’s results (2017), which showed that simultaneous Spanish-Catalan balanced bilinguals accepted in a higher degree the omission of the partitive en than Catalan dominant speakers but they did it in a lesser degree than Spanish dominant, Spanish-Catalan bilinguals. Thus, as Perpiñan (2017)’s results suggested and de Prada Pérez (2009) defended, a language contact situation (also extendible here to bidialectal contact) seems to lead to intermediate outcomes. Overall, and similarly to Vulchanova, Collier, Guijarro-Fuentes and Vulchanov (2022), our results do not seem to correlate with the idea being defended that grammar competence shows irreversible changes in first generation heritage speakers. However, we need to be cautious, since inferential statistics has shown only a marginal significant difference between heritage and monolingual groups of Italian speakers for Condition 3. This could be pointing to a more unstable use of ne for sequential bilingual speakers, which should be confirmed in future studies.

6 Concluding remarks To sum up, our data prove that heritage speakers of Italian mirrored Spanish native speakers showing complete acquisition of both interpretable and uninterpretable features. At first value, our findings seem to be inconsistent with the Interpretability Hypothesis (Hawkins and Hattori 2006; Tsimpli and Dimitrakopoulou 2007), according to which the difficulties that near-natives encounter at the L2 ultimate attainment are related to uninterpretable (i.e., syntactic) features rather than

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 61

interpretable features (i.e., semantic) features. Additionally, it is important to highlight that the Interpretability hypothesis rather than categorical lack of knowledge predicts variability since, according to the authors, L2 speakers tend to develop individual compensation measures. Precisely this source of variation could not be revealed in our dataset. However, considering the successful performance of the heritage Italian group, support for the Interpretability hypothesis seems less clear. On the contrary, considering the Feature Reassembly hypothesis (Lardiere 2009), heritage speakers of Italian were able to acquire the syntactic features needed to omit or to use the object in a L2 similarly to a Spanish native speaker. To wrap up and looking ahead, while interestingly no signs of attrition were found in the long-immersed heritage immigrant community under analysis, the variability found among the Italian monolingual speakers may offer space for future research implying larger groups of monolingual speakers and heritage speakers of second and third generation as well as additional experimental tasks. Our findings, with respect to similar pairs of languages that belong to the same family, suggest that the language attrition is neither feature-specific nor language-specific. It seems that the acquisition of certain features does not depend on the matching of a feature and the target lexical item in question. Thus, future research should consider, for instance, different testing modality, investigating different populations and populations with different types of input. Besides, those linguistic domains more vulnerable to cross-linguistic influence need to be accurately determined since we still do not have a clear-cut picture. Be that as it may, future studies need to continue investigating this line of research to have a clearer and solid picture of potential locus for instability among first generation heritage speakers, since their language constitutes the linguistic input for future heritage speakers.

References Bardovi-Harlig, Kathleen & David Stringer. 2010. Variables in Second Language Attrition: Advancing the state of the art. Studies in Second Language Acquisition 32(1). 1–45. Birdsong, David, LibbyM Gertken & Mark Amengual. 2012. “Bilingual Language Profile: An easy-to-use instrument to assess bilingualism”. COERLL, University of Texas at Austin. Web. 20 Jan. 2012. . Bousquette, Joshua & Michael Putnam. 2020. Redefining language death: Evidence from moribund grammars. Language Learning 70(S1). 185–228 Bruhn de Garavito, Joyce & Pedro Guijarro-Fuentes. 2001. L2 acquisition of indefinite object drop in Spanish. In João Costa & Maria João Freitas (eds.), Proceedings of the GALA 2001 Conference on Language Acquisition, 60–67. Lisbon: Associação Portuguesa de Linguística. Campos, Héctor. 1986. Indefinite object drop. Linguistic Inquiry 17. 354–359.

62 

 Pedro Guijarro Fuentes et al.

Chomsky, Noam. 1995. The Minimalist Program. Cambridge: The Massachusetts Institute of Technology Press. Chomsky, Noam. 2001. On Nature and Language. Cambridge: Cambridge University Press. Clements, J. Clancy. 2006. Null direct objects in Spanish. In J. Clancy Clements & Jiyoung Yoon (eds.), Functional Approaches to Spanish Syntax, 134–150. New York: Palgrave Macmillan. Clifton, Charles Jr., Adrian Staub & Keith Rayner. 2007. Eye movements in reading words and sentences. In Roger P. G. van Gompel, Martin H. Fischer, Wayne S. Murray & Robin L. Hill (eds.), Eye Movements: A Window on Mind and Brain, 341–371. Amsterdam: Elsevier. Cuza, Alejandro & Julio César López Otero. 2016. The acquisition of the semantic values of the Spanish present tense in L2 and heritage Spanish. Revista Española de Lingüística Aplicada/Spanish Journal of Applied Linguistics 29(2). 462–486. de Leeuw, Esther, Monika Schmid & Ineke Mennen. 2010. The effects of contact on native language pronunciation in a migrant context. Bilingualism: Language and Cognition 13. 33–40. de Prada Pérez, Ana. 2009. Subject expression in Minorcan Spanish: Consequences of contact with Catalan. The Pennsylvania State University dissertation. Domínguez, Laura, Glyn Hicks & Roumyna Slabakova. 2019. Terminology choice in generative acquisition research: The case of “incomplete acquisition” in heritage language grammars. Studies in Second Language Acquisition 41(2). 241–255. Duffield, Nigel & Lydia White. 1999. Assessing L2 knowledge of Spanish clitic placement: converging methodologies. Second Language Research 15. 133–160. Gerards, David Paul & Elisabeth Stark. 2021. Why “partitive articles” do not exist in (Old) Spanish. In Tabea Ihsane (ed.), Disentangling Bare Nouns and Nominals Introduced by a Partitive Article, 105–139. Leiden: Brill. Guijarro-Fuentes, Pedro & Cristina Suárez-Gómez. 2021. Tense in speakers of English as a contact language: World Englishes and first generation long-immersed immigrants. English Studies 102(5). 601–622. Gürel, Ayse. 2008. Review of Research on first language attrition of morphosyntax in adult bilinguals. Second Language Research 24(3) 431–449. Håkansson, Gisela. 1995. Syntax and morphology in language attrition: A study of five bilingual expatriate Swedes. International Journal of Applied Linguistics 5(2), 153–171. Hawkins, Roger & Hajime Hattori. 2006. Interpretation of English multiple wh -questions by Japanese speakers: A missing uninterpretable feature account. Second Language Research 22. 269–301. Huang, James. 1984. On the distribution and reference of empty pronouns. Linguistic Inquiry 15. 531–575. Jarvis, Scott. 2019. Lexical attrition. In Monika S. Schmid & Barbara Köpke (eds.), The Oxford Handbook of Language Attrition, 241–250. Oxford: Oxford University Press. JASP Team. 2020. JASP (Version 0.14.1) [Computer software]. Köpke, Barbara & Monika Schmid. 2004. Language Attrition: The next phase. Amsterdam: John Benjamins. Kupisch, Tanja, Dagmar Barton, Giulia Bianchi & Ilse Stangen. 2012. The HABLA-Corpus (German-French and German-Italian). In Thomas Schmidt & Kai Wörner (eds.), Multilingual corpora and multilingual corpus analysis, 163–179. Amsterdam: Benjamins. Lardiere, Donna. 2009. Some thoughts on the contrastive analysis of features in second language acquisition. Second Language Research 25(2). 173–227. Lohndal, Terje, Jason Rothman, Tanja Kupisch & Marit Westergaard. 2019. Heritage language acquisition: What it reveals and why it is important for formal linguistic theories. Language and Linguistic Compass, 13(12).

Chapter 2 Ultimate attainment in long-immersed heritage Italian immigrants 

 63

Mariotti, Alice & Malvina Nissim. 2014. Parting ways with the partitive view: a corpus-based account of the Italian particle ne. Parting Ways with the Partitive View: a Corpus Based Account of the Italian Particle ne. In Roberto Basili, Alessandro Lenci & Bernardo Magnini (eds.), Proceedings of the First Italian Conference on Computational Linguistics CLiC-it 2014 & and of the Fourth International Workshop EVALITA 2014, 249–253. Pisa: Pisa University Press. Masullo, Pascual. 2017. La interfaz sintaxis-pragmática: Caída de objeto acusativo definido sin clítico en el español rioplatense. Saga Revista de Letras 7. 53–72. Montrul, Silvina. 2008. The Incomplete Acquisition: Re-examining the Age Factor. Amsterdam: John Benjamins. Montrul, Silvina. 2016. The Acquisition of Heritage Languages. Cambridge: Cambridge University Press. Montrul, Silvina & Roumyana Slabakova. 2003. Competence similarities between native and nearnative speakers: An investigation of the preterite/imperfect contrast in Spanish. Studies in Second Language Acquisition 25. 351–398. Perpiñán, Silvia. 2017. Catalan-Spanish bilingualism continuum: The expression of non-personal Catalan clitics in the adult grammar of early bilinguals. Linguistic Approaches to Bilingualism 7(5) 477–513. Polinsky, Maria. 2011. Reanalysis in adult heritage language: A case for attrition. Studies in Second Language Acquisition 33. 305–328. Polinsky, Maria. 2018. Heritage Languages and Their Speakers. Cambridge: Cambridge University Press. Robertson, Daniel & Antonella Sorace. 1999. Losing the V2 constraint. In Elaine Klein & Gita Martohardjono (eds.), The Development of Second Language Grammars. A Generative Approach, 317–362. Amsterdam: John Benjamins. Rothman, Jason. 2007. Heritage speaker competence differences, language change, and input type: Inflected infinitives in heritage Brazilian Portuguese. International Journal of Bilingualism 11. 359–89. Rothman, Jason & Michael Iverson. 2013. Islands and objects in L2 Spanish: Do you know the learners who drop? Studies in Second Language Acquisition 35(4). 589–618. Seliger, Herbert & Robert Vago. 1991. The study of first language attrition: An overview. In Herbert Seliger & Robert Vago (eds.), First Language Attrition, 3–16. Cambridge: Cambridge University Press. Schmid, Monika. 2019. The impact of frequency of use and length of residence on L1 Attrition. In Monika Schmid & Barbara Köpke (eds.), The Oxford Handbook of Language Attrition. Oxford: Oxford University Press. Schmid, Monika & Gülsen Yılmaz. 2018. Predictors of language dominance: An integrated analysis of first language attrition and Second Language Acquisition in late bilinguals. Frontiers Psychology 9. 1306. Schmid, Monika Susanne, Barbara Köpke & Kees de Bot. 2013. Language attrition as a complex, non-linear development. International Journal of Bilingualism 17(6). 675–682. Schmid, Monika & Barbara Köpke. 2019. The Oxford Handbook of Language Attrition. Oxford: Oxford University Press. Sorace, Antonella. 2012. Pinning down the concept of “interface” in bilingualism: a reply to peer commentaries. Linguistic Approaches to Bilingualism 2(2). 209–216. Sorace, Antonella. 2005. Selective optionality in language development. In Leonie Cornips & Karen P. Corrigan (eds.), Syntax and Variation: Reconciling the Biological and the Social, 55–80. Amsterdam: John Benjamins. Tsimpli, Ianthi. 2014. Early, late or very late? Timing acquisition and bilingualism. Linguistic Approaches to Bilingualism 4. 283–313.

64 

 Pedro Guijarro Fuentes et al.

Tsimpli, Ianthi, Antonella Sorace, Caroline Heycock & Francesca Filiaci. 2004. First language attrition and syntactic subjects: a study of Greek and Italian near-native speakers of English. International Journal of Bilingualism 8. 257–277. Tsimpli, Ianthi & Maria Dimitrakopoulou. 2007. The Interpretability Hypothesis: Evidence from wh-interrogatives in Second Language Acquisition. Second Language Research 23. 215–242. Vulchanova, Mila, Jacqueline Collier, Pedro Guijarro-Fuentes & Valentin Vulchanov. 2022. Variation in first-generation L1 deictic systems: Language attrition and bilingualism effects. International Journal of Bilingualism. White, Lydia. 2003. Second Language Acquisition and Universal Grammar. Cambridge: Cambridge University Press

Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

Chapter 3  Grammatical competence in adult heritage speakers of Italian and adult immigrants: A comparative study

Abstract: The grammar of bilingual children has been shown to be sensitive to linguistic markers for language impairment. These markers can detect similarities and differences between typical bilingual profiles and atypical monolingual profiles in children. In this chapter, we review a study exploring whether the same markers can detect differences in the grammatical patterns of adult bilingual speakers of Italian immersed in an English-speaking environment. Adult immigrants (AI) and heritage speakers (HS) of Italian are bilinguals who are native speakers of a language that is not dominant in their current environment. The study exploits language markers applied to the investigation of language-specific vulnerabilities in Italian children with language impairments, in particular the production of clitic pronouns and the task of sentence repetition. In both tasks, accuracy in HS is significantly worse than that in AI, showing that both linguistic markers are sensitive to a difference between AI and HS grammatical profiles. In sentence repetition both groups show high accuracy; in clitic production HS are considerably more affected than AI. Qualitatively, the markers show similarities, with most produced sentences being grammatically licit in both groups, but also important differences, with HS showing a selective and more severe disadvantage in the use of functional words modifying sentence structure (complementisers, clitics). Keywords: clitic pronouns, sentence repetition, clinical markers, immigrants, heritage speakers

1 Introduction Within the bilingual spectrum, heritage speakers are speakers who grew up acquiring a language at home which is different from the language of the larger society, as well as the majority language of the host country (Fishman 2001; Rothman 2009; Benmamoun, Montrul, and Polinsky 2013b, Montrul 2008, 2016). Their age of onset of acquisition of the heritage language is similar to that of monolinguals and they are thus considered native speakers of that language. They can be either https://doi.org/10.1515/9783110759587-004

66 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

simultaneous bilinguals, if they are also exposed to the majority language from birth, or successive bilinguals, if contact with the majority language occurs later. Nonetheless,  heritage speakers typically experience a shift in dominance with schooling, when the quantity of input and continued language use shifts from the heritage language (HL) to the societal language (Rothman 2009; Kupisch and Rothman 2018). Shaped by several factors taking place both during the acquisition process and in adulthood, the language competence of these bilinguals on their HL varies considerably and it has become of interest to researchers in all areas of language. One defining factor in childhood is age of first exposure: in grammar, early exposure tends to result in monolingual-like competence at least in some properties of language (Håkansson 1995; Montrul 2010; Montrul et al. 2008; Polinsky 2008). Moreover, heritage speakers who had schooling in the heritage language are generally more aligned to the monolingual standard (Kupisch and Rothman 2018; Dickson et al. 2021). Despite showing differences with the ‘gold standard’, heritage grammars seem to share qualitative similarities in their grammar with monolingual speakers more so than with another natural comparator, namely (adult) L2 learners. It seems to be the case that the two populations -heritage speakers and L2 learners- can have similar accuracy (quantity- see Bianchi 2012 and Romano 2020 and subsequent work discussed below), but differ in their linguistic patterns, where heritage speakers are more aligned to monolingual standards (quality- Montrul 2010; Romano 2020). Importantly, comparisons between heritage speakers and L2 speakers are also claimed to be influenced by methodological considerations. Specifically, L2 speakers are claimed to show an advantage when metalinguistic knowledge is required, whereas heritage speakers to behave more like the native speakers when online tasks are employed (Bowles 2011; Montrul 2016; Montrul et  al. 2008). However, this is no longer the case when heritage speakers have a formal education in the heritage language, suggesting a directly proportional relationship between the level of literacy and metalinguistic knowledge in heritage speakers (Romano and Guijarro-Fuentes, under review). Given the observation that heritage speakers’ profile shares similarities with monolingual native speakers but their attainment of the heritage language is still divergent from ‘native’ grammars, a recent view in bilingual research is to compare heritage languages to the community providing them with the input, rather than to the ‘gold standard’ (Benmamoun, Montrul, and Polinsky 2013a, 2013b; Polinsky 2018, D’Alessandro et al. 2021). Typically, heritage speakers learn their heritage language from long-term immigrants living in a country where their native language is not the majority language. To what extent parents’ input is determining for the quality of heritage grammars is still discussed due to mixed results (e.g., Cuza et al. 2019; Daskalaki et al. 2020), but there is strong consensus in considering the

Chapter 3 Grammatical competence in adult heritage speakers 

 67

variety spoken by long-term immigrants as deviating from the standard variety spoken in the homeland due to heavy exposure to phenomena of crosslinguistic influence (see e.g., Schmid 2011; Polinsky and Scontras 2020 for a review of findings). Crosslinguistic influence is made up of the effects that one language has onto the other, which are described as having one direction, typically from the dominant to the non-dominant language (Yip and Matthews 2000; Paradis 2001; Argyri and Sorace, 2007; Kupisch 2007; Nicoladis 2012). When it is an effect of the dominant, non-native language over the native language, crosslinguistic influence is also referred to as ‘(first language) attrition’ in expats immersed in a dominant language different from their native language, and ‘dominant language transfer’ (Montrul 2016; Polinsky 2018) in conditions of bilingualism. When the interaction between the two languages starts during acquisition (as is typically the case with heritage speakers) phenomena of dominant language transfer are entangled with that of divergent language attainment (Polinsky and Scontras 2020). Crosslinguistic influence appears very early after immersion in an environment where a language different from the native language is spoken (e.g., Linck et al. 2009) and signs of it are manifested in different domains of language i.e., in the lexicon (Kohnert et al. 1999; Köpke 2002 inter alios), at the syntax-discourse interface (Gürel, 2004; Tsimpli et al. 2004; Tsimpli 2007; Sorace 2005, 2011) and more. In syntax, it has been claimed to be a selective process: pervasive enough to influence some surface phenomena (for example overt subject pronouns: Sorace 2011, Chamorro and Sorace 2019, split intransitivity: Montrul 2004), but not strong enough to modify deep structures that would lead to syntactic violations in the standard language (Cuza 2010, 2013; Domínguez 2009, 2013; Gürel 2002; Iverson 2012). The structures vulnerable to crosslinguistic influence have been identified as those affected by contextual or pragmatic conditions and that show a degree of optionality, namely those at the interfaces between syntax and discourse/pragmatics (split intransitives, double object constructions, focus: Sorace 2011; Sorace and Filiaci 2006; Hulk and Muller 2000; Fenyvesi 2005), or those characterised by a complexity in the syntactic dependencies (Wh-dependencies and silent objects: Cuza and Strik 2012; Laleko and Polinsky 2016; Polinsky 2018). Syntactic operations that are acquired late by the monolingual child seem to be more vulnerable to change in a language contact situation and depending on the age of the bilingual (Montrul 2008; Laleko and Polinsky 2016). In this chapter, we will concern ourselves with some aspects of grammatical competence in adult bilingual speakers. Apart from the areas of grammar identified as prone to change in attrition, other areas are shown to be particularly vulnerable in heritage grammars, such as inflectional morphology in the verbal domain, gender, and definiteness agreement in the nominal domain (Polinsky 2008 for Russian and Montrul 2008 for Spanish. See Benmanoun, Montrul, and Polinsky 2013b for

68 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

a review of findings). The study reviewed in this chapter tests grammatical abilities through two linguistic markers for Italian, namely the production of a specific syntactic construction, namely the clitic pronoun, and a psycholinguistics paradigm used across languages for language assessment, namely sentence repetition. In the next sections we will describe the literature exploring linguistic markers for language impairment in bilingualism as well as some studies exploring grammatical properties of Italian in adult bilinguals, before turning to the data in this study.

2 Bilingualism and markers for language impairment In some areas of language, an overlap between (typical) bilingual children and atypical children has been identified, creating the need in clinical practitioners to differentiate atypical monolingual and bilingual from typical monolingual and bilingual profiles, as well as opening up important considerations on the concept of language attainment in bilinguals and the ‘gold standard’. Exploring the overlap is thus of pivotal importance and is an active area of research (Paradis 2010; Armon-Lotem 2015; Vender et al. 2016; Tuller et al. 2018; Blom et al. 2019; Hamann et al. 2020). At the end of the language development, when a language system has reached its final stage (before decline), early assessment is no longer relevant, but observing the language of adult bilinguals is important to define the pockets of grammatical aspects of bilingual populations requiring attention. Moreover, the areas of language which are vulnerable in conditions of language impairment have been shown to highlight critical areas in typical language profiles with different acquisition as well. Therefore, markers of language impairment in the bilingual speaker can be useful tools to explore the language attainment of healthy heritage individuals.

2.1 Clitic production The referential system of some Romance languages, including Italian, features clitic pronominals, as in 1. These are pronominal affixes which appear in special dedicated positions that are not available to other pronouns and noun phrases, and that are typically distinct from the argument position (Kayne 1975). In declarative sentences, they usually appear in preverbal position instead of occupying the canonical postverbal object position. Unlike other referring elements, this position is obligatory. In Italian, clitics are morphologically marked for case and gender in

Chapter 3 Grammatical competence in adult heritage speakers 

 69

the 3rd person, but not in the 1st and 2nd person. In complex verbs, 3rd person clitics also trigger agreement on the participial, as in 2. (1) Il bambino lo bagna the child cl.MASC.PL wet.SG “the child is wetting it/him” (2) Il bambino l(a)’ ha salutata The child cl.FEM.SG has greet.PST.F.SG “the child has greeted her” In monolingual acquisition of Italian, clitic pronouns make an early appearance in spontaneous speech, with children reported to produce clitics at the age of two and to have a fully-fledged system by the age of four (Guasti 1993/4; Leonini 2006a, 2006b; Guasti et al. 2016). Acquisition follows a systematic process whereby children initially undergo a phase of optional omission of the argument, which creates ungrammatical sentences in Italian in the case of the direct object (DO) (Gianni dà, Gianni gives). As production increases, omissions decrease, and they virtually disappear by age 4 (Leonini 2006a, 2006b; Schaeffer 2000). No placement errors are detected in children acquiring Italian, and morphological marking on clitics is in place early, with children performing like adults already at three (Hyams and Schaeffer 2008; Moscati and Tedeschi 2009). Clitics have qualified as an early marker for Developmental Language Disorders (DLD) in clitic languages (French: Jakubowicz et al. 1998; Hamann et al. 2003; Tuller et al. 2011; Italian: Bortolini et al. 2002, 2006; Arosio et al. 2010, 2014 inter alios). Children with a diagnosis of DLD are significantly less likely to produce the target object clitic (26% of target clitics produced in the DLD group, 96% in the age-matched controls in Bortolini et al. 2002, 2006). Frequent non-target answers contain omissions in younger children (Bortolini et al. 2002, 2006; Leonard and Dispaldro 2013 for Italian, Hamann et al. 2003 for French), and the production of lexical NPs in older participants (Hamann et al. 2003, French, Arosio et al. 2014; Guasti, et al. 2016, Italian) all the way into early adulthood (French, 11;5–20;5, Tuller et al. 2011). While specifically a marker for clinical conditions, the production of clitic pronouns is a sensitive area in any language system where attainment of syntax differs from standard, both in acquisition and at the end of language development. This is the case for typically-developing bilingual children, where spontaneous production shows a similar pattern to that of monolingual acquisition showing an optional omission stage (Ferrari 2006; Serratrice, Sorace, and Paoli 2004 on bilingual children with a Germanic L1), but productions are lower: the monolinguals in Leonini

70 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

(2006a, 2006b) correctly produce around 86% of clitics in expected contexts, while at the same age Vincenzo and Elisa (Ferrari 2006) produce 57% and 59%, respectively. Bilingual children resort to the production of lexical noun phrases more frequently than their monolingual counterparts when their other L1 is a non-clitic language (Paradis 2004; Rogers 2009; White 1996; Belletti and Hamann 2004), and errors on the clitic itself (of gender, number, both gender and number, or case) are more frequent when children come from a clitic L1 (Vender et al. 2016). Errors of misplacement, just like in monolingual typical acquisition, have not been attested in Italian bilinguals, but they have in French (Adiv 1984; Granfeldt and Schlyter 2004). When Italian is learnt as an L2, the supply of clitics in elicited contexts is low (39% of elicited clitics in Leonini and Belletti 2004, 28% in Leonini 2006a, 2006b for children coming from a Germanic L1). Alternative structures include omissions and lexical NPs. In L2 Italian, some studies report placement errors on the clitic (Granfeldt and Schlyter 2004), while others do not (Leonini and Belletti 2004).

2.2 Sentence repetition Sentence repetition is a psycholinguistics paradigm which has been shown to be sensitive to atypical linguistic profiles including DLD, Autism Spectrum Disorder (ASD), agrammatic aphasia, as well as in typical acquisition (Lust et al. 1996; Friedmann and Grodzinsky 1997; Friedmann 2001, 2007; Friedmann and Lavi 2006; Fattal et al. 2011, Sukenik and Friedmann 2018). Its power as a diagnostic tool lies in the fact that repetition of a sentence is not simple phonological reiteration: repetition of a string that exceeds the word span of an individual has been claimed to be possible because the individual reconstructs the stimulus with information from long-term memory, specifically anchoring it to lexical, conceptual, and syntactic representations (Clay 1971, Potter and Lombardi 1990, 1998, Lombardi and Potter 1992). It follows that the paradigm shows a participant’s ability to process different grammatical structures. Riches and colleagues (Riches et al. 2010) designed a sentence repetition task for English targeting different levels of syntactic complexity by combining two factors, namely relative clause type (subject and object) and adjective position (in main clause vs in relative clause) and tested the clinical groups of DLD and ASD with language impairment (ALI), and a typical group of adolescents. Results showed that, while both clinical groups showed higher error rates than the control group, complexity had a more prominent effect in DLD than it did in ALI: participants from the DLD group showed selectively more errors in the more complex structures, and the ALI group showed across-the-board errors. Similar results for the two groups in a sentence repetition task were obtained in Sukenik & Friedmann

Chapter 3 Grammatical competence in adult heritage speakers 

 71

(2018) for Hebrew, testing sentences involving different types of syntactic movement. A delayed sentence repetition targeting tense and agreement was carried out for agrammatic aphasia in Friedmann and colleagues for Hebrew (Friedmann and Grodzinsky 1997, Friedmann 1998), which showed that while verbal, adjectival, and nominal agreement are not affected in this population, tense marking is. This subtle distinction allowed for deductions on the nature of the deficit, which, though targeting morphology, lies in the distinction between different inflection operations. The language of children with DLD has been shown to share some properties with that of sequential/early bilinguals as tested through sentence repetition tasks. In Meir and colleagues (2015), the Russian and Hebrew adaptations of the LITMUS-SRep task (Marinis and Armon-Lotem 2015) were administered to two groups of Russian-Hebrew bilingual children, one with typical development and one with DLD, and two groups of monolingual children for each language. The task tested morphology and syntactic structures of increasing complexity, from active sentences with canonical order to biclausal sentences. SRep tasks in both languages involved grammatical morphology as well as complex syntactic structures reported to be difficult for children with DLD, and sentences were grouped into three levels of complexity specific to each language as determined by state-of-the-art results, from simple sentences with no dependencies to sentences with dependencies. Results showed that sentence repetition was a valuable tool to discriminate children with DLD even among bilingual children. In fact, while bilingual children showed overall lower results than the monolingual typically developing (TD) children, bilingual children with DLD performed significantly worse than their bilingual TD peers. Moreover, distinctions were not only limited to accuracy: sentences with higher complexity were more problematic in DLD than in TD bilinguals, and bilingual children with DLD produced error patterns comparable to those reported for monolingual DLD, and not TD bilinguals. Meir et al.’s results highlight the pivotal role of sentence repetition in the description of different language systems: both accuracy on sentence type and qualitative information given by alternative answer analysis can be informative and show differences between populations.

3 Italian syntactic competence in heritage speakers Italian has been at the heart of some important investigations on grammar in crosslinguistic influence, particularly in conditions of attrition. Studies from Sorace and colleagues (Tsimpli et al. 2004, Sorace 2011) on Italian long-time expats who were

72 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

proficient L2 speakers of English have highlighted several aspects of grammar where their language differs from that of monolinguals living immersed in Italian in the homeland. In Tsimpli et al. (2004), participants were tested with a pictureidentification task on the interpretation of the referent of overt and null subject pronouns in contexts of backward (3b) and forward anaphora (3a). Reference assignment varies according to the null or overt nature of the subject, and its position in the clause. In null subject languages like Italian, null subjects (pro in the examples below) are the default, non-marked alternative, while overt subject pronouns are marked. When the antecedent is in the following sentence, like in 3a, if the subject is null then coindexation between pro and the matrix subject is the default. On the other hand, if the subject is overt, coindexation with the matrix subject is not accepted in the syntax, and the referent is either the object of the matrix clause or a third referent. In 3b, the overt subject pronoun cannot be assigned the same referent as the subject of the matrix sentence while pro is ambiguous between coindexation with the subject or the object of the matrix clause. (3) a. Quando leik/l / proi attraversa la strada, l’anziana signorai saluta la ragazzak. ‘When she crosses the street, the old lady greets the girl’ b. L’anziana signorai saluta la ragazzak quando leik/l / proi attraversa la strada ‘The old lady greets the girl when she crosses the street’ Italian (as well as Greek) speakers in the L2 setting showed no effect of attrition on reference assignment when this is governed by structural features (sometimes referred to in the literature as uninterpretable features) both of null subjects, as in 3a, and of overt subjects, as in 3b. On the other hand, they significantly differed from the monolinguals in the control group in their interpretation of overt pronouns in 3b, when choice is not governed by syntax. The authors concluded that attrition may affect linguistic phenomena that show optionality due to their collocation at the syntax-pragmatics interface, but not when interpretation does not allow for optionality in the native grammars. Adult heritage speakers as well as L2 speakers of Italian were tested on their abstract representations of clitics in a structural priming task by Romano (2020, 2021), which focused on three positional differences featuring different verb types (lexical, causative, and modal) as in 4. The sentences are interpreted by the author as being on a scale of complexity, with lexical verbs originating the least complex sentences (with no dependencies), and modals originating the most complex sentences (with dependency and clitic climbing).

Chapter 3 Grammatical competence in adult heritage speakers 

(4) a. Modal

b. Causative

c.

Lexical

 73

i pesci, Pietro li vuole cucinare the fish.PL Pietro cl.MASC.PL want.3SG cook.INF all’ aperto at-the open “The fish, Pietro wants to cook them outdoors” i pesci, Pietro li fa cucinare the fish.PL Pietro cl.MASC.PL make.3SG cook.INF all’ aperto dalla Zia at-the open by-the Aunt “The fish, Pietro has them cooked outdoors by the aunt” i pesci, Pietro li cucina the fish.PL Pietro cl.MASC.PL cook.3SG “The fish, Pietro cooks them outdoors”

all’aperto at the open

In the structural priming task, participants saw a picture containing a priming sentence that they were instructed to read out loud. Then, they saw a picture with four prompts eliciting the targeted structure and were asked to form a sentence. Participants were speakers of Swedish and Italian living in Sweden. In his studies, Romano finds that heritage speakers were less accurate than the monolingual speakers in the production of a primed clitic pronoun. In terms of accuracy, their performance was similar to that of the L2 speakers. This result was replicated in a truth-value judgement task featuring similar grammatical constructions. Illicit structures produced in the structural priming task contained mostly omissions of the clitic. However, a similarity between heritage speakers and monolingual speakers of Italian was found on the effects of priming: the type of elicited verbal construction influenced the effect of priming in a similar way in heritage speakers and monolingual speakers, with both exhibiting stronger priming effects in lexical and causative constructions over modal constructions. On the other hand, L2 participants exhibit a lexical>causative>modal pattern. The author concludes that while the language of heritage speakers and that of monolingual speakers bears some substantial differences in accuracy and in the number of non-standard productions, abstract representations are qualitatively similar between heritage speakers and monolingual speakers. The same design was implemented in Romano (this volume) to test attainment of gender in heritage speakers of Italian with dominant Swedish. Gender is intended as both the morphological feature assigning an object or person either masculine or feminine (in Italian), and the syntactic operation of agreement,

74 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

whereby e.g., a feminine noun triggers agreement with other constituents of the sentence, such as the determiner and the past participle, as in 5. Lexically, both languages have declension classes, but the Swedish system is opaque, while the Italian one is not. (5) La bambina biond-a the.SG.FEM girl.SG.FEM blonde.SG.FEM “the blonde girl” In Italian, gender agreement must also appear on the clitic pronoun where, contrary to other types of pronominals like the ones employed in Swedish, it is an uninterpretable feature and is thus reliant on a syntactic operation. Data from the study show heritage speakers to be less proficient in producing clitics whose features correctly agreed with those expressed on the dislocated NP that was their referent, as in 4 above. However, the difficulty did not seem to lie on the operation of gender agreement (namely on the production of the correct features as such): in fact, the most frequent alternative structure produced was the (illicit) omission of the clitic, with only a handful of errors on gender features. The differences in the grammars of heritage speakers and monolingual speakers were replicated in the (timed) truth-value judgement task, but again the gender features of the clitic did not seem to have a role in accuracy, as no statistical significance was found between masculine and feminine clitics. Adult bilingual speakers were tested on Italian gender also in Bianchi (2012). In this case, the majority language was German. Both Italian and German express lexical gender and gender agreement. While Italian has two lexical genders (feminine and masculine) and these are generally transparent, German has three (neuter) and, like Swedish, the categorisation is less transparent. Both languages show agreement of the features between the noun and the determiner, while Italian also shows agreement on the past participle, as in 6. (6) La bambina è andat-a the.SG.FEM girl.SG.FEM is gone.SG.FEM “the girl went to the park”

al parco to-the park

The participants of this study were highly proficient L2 learners of Italian with L1 German (defined as L2-Italian), Italian-German bilinguals living in Italy (defined as 2L1-Italian strong by the author), and Italian-German bilinguals living in Germany (defined as 2L1-Italian weak). The latter group is what we would now refer to as heritage speakers of Italian. In an acceptability judgement task, participants had to repeat or correct a sentence they heard containing an AdjP as well as a clitic. In

Chapter 3 Grammatical competence in adult heritage speakers 

 75

7, a masculine NP (pettine, comb) does not agree with its determiner and the past participle. (7)

Ho usato la pettine verde pro have.3SG used.PST the.SG.FEM comb.SG.MASC green e l’ ho rimess-a nel cassetto and cl.SG.FEM have.3SG put-back.PST.SG.FEM in-the drawer ‘I used the green comb and I put it back in the drawer.’ ✶

In an elicited production task, sentences containing a lexical NP and a clitic matched in gender were elicited. Bianchi found that the heritage speakers of Italian deviated from the target more than the bilinguals living in Italy. Much like in Romano’s study (ibidem), the domain of vulnerability in the heritage language was not the operation of gender agreement, but rather the lexical aspect of gender assignment (namely deciding whether an Italian noun was masculine or feminine). Similar results were reported for the L2 group. For the purposes of this discussions, we can conclude that the heritage speakers did not have issues with the syntactic operation of agreement, as is expected from them if syntax is to be considered intact. In the studies on the syntax of heritage speakers of Italian reviewed above, Italian as a heritage language (at least for speakers of a Germanic language) is shown to share similarities with the Italian acquired as an L2 in adult life, as demonstrated by the fact that heritage speakers are less accurate on some grammatical elements such as gender morphology and clitic pronouns in complex constructions. However, it also shares traits with the Italian spoken in the standard variety in some of its qualitative patterns. Accuracy was mostly vulnerable in surface grammar, as demonstrated by the lower impact on structural operations such as gender assignment, or complex constructions, as demonstrated by the omissions in clitic climbing, whereas structure (or uninterpretable features) seems relatively spared.

4 Current study An objective of current studies on bilingualism is to give formal descriptions of the languages spoken by (different) bilingual populations in their final state. This is of paramount importance both on theoretical grounds and for differentiating the bilingual language profile from an atypical one. On theoretical grounds, gaining better understanding of what undergoes change in bilingual contexts (and how or to what extent depending on the population) helps us disentangle the phenomenon

76 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

of crosslinguistic influence and gives us indications of what to expect from the performance of specific subgroups of bilinguals. Moreover, crosslinguistic influence is proposed to foreshadow diachronic change in a shorter period of time than in monolingual settings, creating a window on what might happen to language over time (Rinke and Flores 2014; Flores and Rinke 2020; Nagy 2016, 2017; D’Alessandro 2021). For clinical purposes, it is important to give clinicians precise indications on what to expect from the language attainment of bilingual populations, should assessments be required later in life due to degenerative conditions or late diagnoses. As discussed in the previous sections, linguistic markers identified for language disorders, both in the form of language-specific elements such as tense and agreement marking for English or clitics for French and Italian, and in the form of paradigms such as nonword and sentence repetition, may also be vulnerable in typical bilingual populations, although qualitative differences may occur. The purpose of this study is to test whether linguistic markers identified for Italian language disorders, namely the production of clitic pronouns and the task of sentence repetition, are also sensitive to situations of normal bilingualism where Italian is not the dominant language (RQ1). Furthermore, it aims to give further indications on how (Italian) heritage grammars differ from the languages of expat communities (RQ2). Given that these tasks can be vulnerable in conditions of multilingualism, such as in adult L2 learners and successive or sequential bilingual children, we hypothesised that accuracy for the two tasks would highlight differences between heritage speakers and their baseline, with heritage speakers showing lower accuracy on targeted structures (H1). At the same time, the two populations featured in the study are native speakers of the language under investigation. For this reason, the answer patterns of the two populations were predicted to be qualitatively similar, and to not feature structurally illicit sentences (H2).

5 Methods 5.1  Participants A total of 59 adult participants took part in the study. These were divided in two groups according to whether they were native speakers of one or both the languages in their environment: 30 were native speakers of English and Italian born in the UK from a first-generation Italian family, referred to as heritage speakers (HS), mean age 35 years; 29 were first-generation expat native speakers of Italian, referred to as adult immigrants (AI), with a mean age of 39 years. Only two were the parents of

Chapter 3 Grammatical competence in adult heritage speakers 

 77

heritage speakers in our group, although the others shared a similar immigrational background as first-generation immigrants as the other heritage speakers’ families of origin. All participants were living in Scotland at the time of testing and had been living in the UK most of their life (all heritage speakers were born there except two, who had moved before primary school). 26 out of the 29 heritage speakers had both parents who were native speakers of Italian, while the remaining three had one of the two. Literacy in Italian was also consistent in this group, with all participants having had no schooling in Italian. Adult immigrants were also consistent in that they were all native speakers of Italian and were consistently exposed to English only after moving to the UK as adults (>20 years). All AI were formally educated in Italian; however, six of them completed their higher education in the UK. All participants performed a language profile questionnaire, LEAP- Q (Marian, Blumenfeld, and Kaushanskaya 2007) detailing their exposure to both languages throughout their lifetime and at the time of testing. Table 1 summarises measures of the LEAP-Q. Table 1: Descriptive data for the two groups, Adult Immigrants (AI) and Heritage speakers (HS): mean age in years (and SD), Age of first Exposure to Italian (AoE), level of education, mean years of formal education in Italian (and SD), mean years in the UK (and SD). Age (SD)

AoE

Education

Formal education in Italian

Years in the UK

Adult immigrants (n= 29, female 18)

39.31(11.76)

birth

Higher Edu: 29

16.18 (2.50)

15.25 (8.92)

Heritage speakers (n= 30, female 19)

35.7 (12.29)

birth

Secondary: 10, Higher Edu: 20

0.04 (0.19)

35.4 (11.98)

5.2 Materials All participants completed two background tasks and three experimental tasks.

5.2.1 Background tasks The background tasks measured participants’ competence in comprehension of both Italian and English. For Italian, a reduced online version of the standardised sentence to picture matching task Comprendo (Cecchetto et al. 2012) was adapted on PsychoPy (Peirce et al. 2019) for this study, where accuracy was measured. Upon

78 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

hearing a sentence, participants were instructed to press one of two keys (x, n) mirroring two images shown on the screen, depending on which image best represented the sentence. A total of 30 items were selected from the long version from the following categories (six items per category): actives, coordinations, passives, subject relatives, and object relatives. Items were randomized for each participant. Examples of the selected items are given in Table 2. Table 2: Examples for each type of sentence in the Comprendo task (Cecchetto et al. 2012). SENTENCE

TYPE

La mamma sta baciando il bambino “The mother is kissing the boy” 

Active

Il gatto viene morsicato dal cane “The cat is being bitten by the dog”

Passive

La bambina che tira il cane guarda il bambino “The girl who is pulling the dog is looking at the boy”

Subject Relative

L’uomo che il bambino guarda mangia la torta “The man that the boy is looking at eats the cake”

Object Relative

La bambina mangia la torta, e il bambino beve il latte “The girl is eating the cake, and the boy is drinking milk”

Coordination

Comprehension of English was tested through the standardised task TROG-2 (Bishop 2003). Like the Italian test Comprendo, TROG-2 is a sentence to picture matching task comprising a range of constructions for a total of 80 items. Participants listen to a set of pre-recorded sentences as they watch four images on a computer screen. Each image is labelled a, b, c, or d. Participants are asked to name the letter of the picture the sentence corresponds to. Unlike Comprendo, where items -and consequently the type of sentence- are randomised, the items in TROG-2 are presented in a fixed order of increasing complexity. Both groups reported high scores in the two background tasks. A mixed effects logistic regression showed that in English comprehension the groups perform differently, with adult immigrants scoring an average 1.49 points less in the English task than heritage speakers. This is to be expected considering that while HSs are native speakers of English, AIs are (proficient) L2 speakers. On the other hand, in the comprehension of the language of testing, namely Italian, the two populations behave similarly, in line with accounts of heritage languages as showing high proficiency in comprehension, measured in comparison with standard tests. Scores on the accuracy for both tests are reported in Table 3.

Chapter 3 Grammatical competence in adult heritage speakers 

 79

Table 3: Raw scores (and SDs) for each group in the background tasks Comprendo (Italian) and TROG-2 (English). Comprendo correct (SD)

TROG-2 correct (SD)

Adult Immigrants

28.7/30 (0.2)

77.6/80 (2.02)

Heritage Speakers

28.1/30 (0.22)

79.1/80 (0.99)

5.2.2 Experimental tasks For the purposes of this study, participants were tested on the following language markers for Italian: the production of DO clitics and a sentence repetition task featuring different sentence types. Participants’ processing of Italian sentences was tested through a sentence repetition task developed for Italian (FAST, Di Domenico et al. in preparation). The task is designed to be a quick evaluation of Italian structure processing through a comprehensive list of Italian constructions. The sentences are designed to target structure complexity and/or sentence length, for a total of 26 sentences. The complexity of the structure is determined by whether the item contains a syntactic operation which changes the order of the constituents. Participants are instructed to listen to the sentence read by the researcher and repeat it immediately after. Responses are scored as 1 if the repeated sentence is identical, 0 if it contains an omission or a substitution or if it is not completed. A production was considered a substitution whether it was lexical, namely if a lexical word was substituted with another lexical word (i.e., papà, dad > nonno, granddad), morphosyntactic, if a morpheme was substituted with another morpheme (i.e., va, he/she goes > vanno, they go) or a function word was substituted with another function word which changed the structure of the sentence (la mamma e il nonno, mum and granddad > la mamma con il nonno, mum with granddad). Phonological substitutions that did not change the meaning of the word were disregarded. Sentences are presented in a fixed order. Table 4 gives examples of the items and the corresponding sentence type and locus of difficulty. Production of DO clitic pronouns was examined through a short version of the elicitation task by Arosio et al. (2014). In the task, participants were shown sets of two pictures involving a character and an animate or inanimate object. In the first picture, a sentence introduced the character of the story and told that he/she wanted to perform an action on the object. In the following picture, participants were asked to answer a question about what the character did. The sentences were recorded by a native speaker of Italian and played through loudspeakers.

bianco white

è is.3SG

il gatto che i cani The cat.MASC.SG that the dog.MASC.PL “The cat that the dogs are chasing is white”

(d)

inseguono chase.3PL

La mamma guarda la televisione e the mum watch.3SG the television and “Mum is watching television, and dad is reading the newspaper”

(c)

il giornale the newspaper

il papà the dad

accarezza stroke.1SG

la nonna la The grandma herCL.FEM.SG “Grandma is stroking her”

(b)

legge read.3SG

short sentence, no structural complexity

Il gatto è the cat.MASC is “The cat is black”

long sentence, structural complexity

long sentence, no structural complexity

short sentence, structural complexity

Description nero black. MASC

Sentence (glossa, translation)

(a)

omission of complementiser, reversibility

length

no predicted error

no predicted error

Predicted error

Table 4: Examples of the sentences contained in the sentence repetition task, the description of the sentence in terms of length and complexity, and the predicted error.

80   Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

Chapter 3 Grammatical competence in adult heritage speakers 

 81

(8) Preamble

In questa storia, una signora vuole dipingere una maschera. ‘In this story, a lady wants to paint a mask.’ Probe Guarda, cosa fa la signora alla maschera? ‘Look, what is the lady doing with the mask?’ Elicited answer la dipinge pro it.CL.FEM paint.3SG ‘she painted it’

The task contained a total of 7 items eliciting the DO clitic. All elicited DO clitics were singular (4 masculine and 3 feminine). All probes elicited a present simple, but answers were felicitous both with a present and with a past tense. Three familiarisation trials were given at the beginning of the session. Because the task was originally adapted following Tedeschi (2009) to also test use of the pragmatic context to determine the correct referential expression to be used (namely the clitic or the R-Expression) (Smith 2021), the 7 items eliciting a clitic were alternated with 7 further items targeting an R-Expression. In these items, the PP was not repeated in the probe and was therefore meant to be treated as new information in the elicited answer. For the purposes of this chapter and given that it would not be expected of healthy adult populations for the NP condition to pose any issue, only data from the clitic condition will be discussed.

5.2.3 Procedure Each participant was individually tested in a quiet room on the researcher’s laptop, or via a Zoom call where the participant was required to be in a quiet room, have a large screen, and a headset or loudspeakers; the experimental protocol was administered in one session lasting about 40 minutes. The clitic task was run on PowerPoint where images in colour were shown, and a recorded voice of a female Italian native speakers from northern Italy played on loudspeakers enunciated the sentences. Sentences for the sentence repetition task were enunciated during testing by the authors, who are native speakers of Italian. Each session was recorded, and all materials were transcribed by the authors. The study was approved by the University of Edinburgh ethics committee (ethics application number 35–1920/4).

6 Results Accuracy on sentence repetition and clitic production of the two groups is presented in Table 5. Statistical analyses were run in R (R Core Team 2020).

82 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

Table 5: Mean raw scores of responses (and SDs) for each group in the experimental tasks of sentence repetition and clitic production. Sentence Repetition

DO clitic production

Adults Immigrants

25.8/26 (0.06)

5.6/7 (0.40)

Heritage Speakers

24.8/26 (1.74)

2.5/7 (0.48)

6.1 Sentence repetition A linear model was run to predict the score in sentence repetition depending on the group. Although statistical significance between the two groups is reached R2 = .02, F (1,1584) = 37.95, p = .001, with AIs scoring on average 1 point more than HSs, mean accuracy was very high for both groups (≥95%). The distribution of scores visualised through the density plot reported in Figure 1 shows that the AIs’ scores are mostly clustered in the highest score (26 in the x axis). HSs’ scores are also clustered between 25 and 26, but the distribution is more uneven with several participants also obtaining lower scores. Next, participants’ utterances were analysed for types of errors. Syntactic errors were predicted on the basis of the types of structures featured in each sentence, as shown in Table 4 above. The most common mistakes made by HS were partial or full omissions. Most omissions were of the clitic pronoun, which resulted in an illicit sentence with a dropped object (9). Partial omissions consisted mostly in the dropping the IO clitic in the clitic cluster (10). This sentence is structurally licit. (9) Target:

Produced:

(10) Target:

la mamma la lava the.SG.FEM mum her.CL.FEM wash.3SG ‘the mother is washing her’ ✶ La mamma lava the.SG.FEM mother wash.3SG ‘the mother washes’

Maria glielo paga Maria him.CL.IO-it.CL.DO.MASC pay.3SG ‘Maria pays it for him’

Produced: Maria lo paga Maria it.CL.DO.MASC ‘Maria pays it’

Chapter 3 Grammatical competence in adult heritage speakers 

 83

Density

1.5

1.0

Group heritage immigrant

0.5

0.0 20

22

24

26

Sentence repetition score Figure 1: Density plot of the scores on sentence repetition by group.

Another common mistake among HS was the substitution of the complementizer che with the coordinating element e, which results in transforming a sentence with a dependency, namely a relative clause, into a sentence with no dependency (11), which is still linearly long but does not require building an embedded sentence. (11) Target: Produced:

la nonna guarda il cane che la bambina spinge ‘Grandma watches the dog that the child pushes’ la nonna guarda il cane e la bambina spinge ‘Grandma watches the dog and the child pushes’

84 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

Another type of mistake among HS participants was agreement, between the gender of the determiner and that of the noun (12a), and between the NP and the verb (12b): (12) a. Target:

[. . .] la nonna prepara la grandma.FEM prepare.3SG the.SG.FEM cena dinner.FEM ‘grandma prepares dinner’ Produced: ✶[. . .] la nonna prepara il the.SG.FEM grandma.FEM prepare.3SG the.SG.MASC cena dinner.FEM ‘grandma prepares dinner’ b. Target: il gatto che il cane The.SG.MASC cat.SG that the.SG.MASC dog.SG insegue follow.3SG ‘the cat that the dog is following’ Produced: ✶il gatto che il cane the.SG.MASC cat.SG that the SG.MASC dog.SG inseguono follow.3PL ‘the cat that the dog are following’

The very few errors made by AIs were lexical omissions (e.g., of an adjective or PP, which maintain the sentence licit) or lexical substitutions.

6.2 DO clitic production Mean accuracy in production or 3rd person DO clitics is visualized in Figure 2 below. A  binomial mixed effects logistic regression was run where Score was predicted by the fixed effect of group, type (clitics, R-expression), and their interaction, with random intercept and slope for type by subject and random intercept by item. The linear effect of group was significant, z = 2.06, p = .04, type had an effect where object clitics are significantly harder than R-expressions, z = -3.87, p < .001, and there is a significant interaction between group and type z = 2.49, p = .01. In the clitic condition, HSs perform considerably lower than AIs. 61% of the time, HSs produce NPs in place of the elicited clitic. Whilst not the target answer, this answer is struc-

Chapter 3 Grammatical competence in adult heritage speakers 

 85

turally licit. Only 13 answers (4%) contain a clitic with a feature error (namely the wrong gender or number) or the wrong argument error (namely IO in place of DO), and 9 (3%) contain the use of the full pronoun, and only six omissions were reported. In the same condition, AIs produce a lexical NP 23% of the time. In the lexical NP condition, in all cases in which AIs do not produce the target lexical NP (14% of the time) they produce the corresponding (correct) clitic. The same is true for most alternative answers provided by HSs (7%), save for 3 irrelevant answers.

100%

% of correct answers

75%

Group immigrant

50%

heritage

25%

0% immigrant

heritage

Group Figure 2: Bar plot representing percentage of correct answers on production of DO clitic by group.

86 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

7 Discussion Background measures on comprehension of both simple and complex sentences in Italian showed that both Heritage speakers of Italian and Adults immigrants have high accuracy in the language non dominant in their environment. Thus, abstract representations of Italian grammar in heritage speakers seem to be similar to those of speakers of Italian who were raised in an Italian-dominant environment, as they resemble both the performance of their (bilingual) baseline, and the adult monolinguals (Cecchetto et al. 2012). Regarding our first research question (RQ1), investigating whether language markers identified for Italian language disorders are sensitive to bilingual grammars in adulthood, data from the experimental tasks show that the two markers tested in this study are informative on the language of the two populations and the differences that occur between them. These markers, testing computation during production of Italian and production of a specific construction which is not present in the dominant language, both highlight a difference in the grammars of heritage speakers and their baseline. H1, predicting a difference between accuracy of the two groups, is borne out, as is the predicted direction: in both tasks, HS are significantly less accurate than their baseline. This is true of both sentence repetition and clitic production, but it is particularly relevant for the latter. In sentence repetition, in fact, both groups perform correctly in ≥95% of the cases, and therefore the statistical difference may be considered negligible (Brown 1973); in elicited production of the clitic pronoun the difference is wider: AIs have high accuracy (around 85%) on this element which is learnt early in typical monolingual development and is employed with a strong preference over any other referring expression in Italian, while HSs do not (around 35%). This result is in line with the other studies on production of primed clitics reviewed before (Romano 2020, 2021). Interestingly, while productions of clitics are low in HSs, these are produced roughly 1/3 of the time they are elicited. This may be the case because they are highly available in their input language. In fact, when these are not as highly accessible in the input language, as was shown to be the case for double object clitics, productions in HSs are either much lower or completely absent (Spelorzi et al. 2022). The fact that differences would emerge in the production tasks but not in the comprehension task is in line with some results on heritage syntax in different modalities, where even when a structure was generally absent in the spoken language of heritage speakers, abstract knowledge of the same structure was present as tested through comprehension tasks (for example comprehension of

Chapter 3 Grammatical competence in adult heritage speakers 

 87

German passive constructions in heritage speakers of a dialect of German spoken in Kansas, Putnam and Salmons 2013). Similarly, the fact that the most severe detachment of heritage grammar from standard Italian is noticeable in the elicitation task rather than in the sentence repetition task (where clitics were also featured) could be dependent on the task, which allows speakers for more freedom in the choice of structure than a repetition task. To answer the second research question of this study (RQ2), investigating how (Italian) heritage grammars differ from the grammars of Italian expat communities, we looked at answer patterns. As we have anticipated, while most AIs produce clitics in some if not all the items eliciting clitics, the large majority of HSs always produces lexical NPs. The production of the construction containing this type of pronoun is thus preferred in adult immigrants over the production of (another pronoun or) a lexical NP when the context calls for it, but it is dispreferred in heritage speakers in the same contexts. Nonetheless, structural errors are few in heritage speakers, and none in adult immigrants. In the elicitation task, errors of illicit omissions and misplacements are not recorded, and just a few illicit omissions are recorded in the sentence repetition task. Similarly, few assignment errors were found (of gender and/or number) in both tasks. Another, similar mistake that the heritage speakers sometimes make is substituting the complementiser che with the coordinating element e. Both phenomena indicate structure simplification. In fact, sentences containing a clitic structurally require a movement operation which results in non-canonical argument order in the linear structure, as well as requiring other operations such as agreement. When the participant either produces a structure that both maintains the canonical order and requires no further operations -namely the production of a lexical NP- or, when avoidance is noy an option due to the task requirements, drops the clitic, they are constructing a simpler sentence. Similarly, substitution of the complementiser che for e takes the structure from one with a dependency to one with no dependency. H2, in which we hypothesised that the quality of the grammar of the two populations was similar, is partially borne out: the syntax of both heritage speakers and adult immigrants is (mostly) canonical; however, while adult immigrants make no structural mistakes, heritage speakers are particularly vulnerable in the production of clitics, an issue which is resolved through the use of a lexical NP in the elicitation task, creating a licit sentence, but with partial or total omissions of the clitic in the sentence repetition task. Taken together, these results suggest that while core syntax is overall in place both in abstract comprehension and in production in heritage speakers, sentence complexity may play a role in answer strategies, and higher complexity can lead to errors on core syntax as well.

88 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

8 Conclusions In this study, heritage speakers of Italian who grew up bilingual and are immersed in an environment where Italian is not the dominant language were compared to expat Italian speakers who grew up monolingual Italian and were later immersed in the same environment. Their grammars have been investigated in areas of language reported as vulnerable in both late and early second language learners and atypical populations. A difference between the two groups emerged in accuracy, with adult immigrants reporting a more consistent use of clitic pronouns and no mistakes in sentence repetition, and heritage speakers showing little use of clitic pronouns favouring lexical NPs instead, and a higher number of mistakes in sentence repetition. Qualitatively, adult immigrants produced no structural or morphosyntactic errors across tasks. Lower competence in specific grammatical computations was evident in the heritage group. While they resorted to licit alternative structures in most cases, some non-canonical sentences were produced, particularly in sentence repetition on sentences featuring clitics, complementisers, and gender agreement operations. Taken together, the data presented support the idea that the bilingual experience leads to a continuous process of language change. Bilingual speakers who learn a language from a population of native speakers outside of their homeland seemingly acquire a language that shows changes even from the language of their input. This is not only evident at the interface between linguistic modules, but also in specific areas of grammar such as pronominal clitics or complex sentences featuring, in the present case, complementisers. This result goes in the direction of showing that differences can also be found in structural language. Importantly, heritage speakers only produced illicit sentences in SR but not in the elicitation task, suggesting that, when allowed to create novel sentences using their preferred grammars, structural rules of the target language are followed, and changes in the language from a potentially attrited population to a heritage one may be more in terms of preference. More studies on different bilingual experiences are required to better understand factors that modulate the grammatical competence of speakers with different opportunities to practice the language. These studies are pivotal to understanding the phenomena at the core of language change in both bilingual and monolingual settings.

Chapter 3 Grammatical competence in adult heritage speakers 

 89

References Adiv, Ellen. 1984. Language learning strategies: the relationship between L1 operating principles and language transfer in L2 development. In Roger Andersen (ed.), Second languages: A crosslinguistic perspective, 125–142. Rowley, MA: Newbury House. Argyri, Efrosyni & Antonella Sorace. 2007. Crosslinguistic influence and language dominance in older bilingual children. Bilingualism: Language and Cognition 10(1). 79–99. Armon-Lotem, Sharon, Jan de Jong & Natalia Meir (eds.). 2015. Assessing Multilingual Children. Disentangling Bilingualism from Language Impairment. Bristol: Multilingual Matters. Arosio, Fabrizio, Chiara Branchini, Luca Barbieri & Maria T. Guasti. 2014. Failure to produce direct object clitic pronouns as a clinical marker of SLI in school-aged Italian speaking children. Clinical Linguistics & Phonetics 28. 639–663. Arosio, Fabrizio, Chiara Branchini, Matteo Forgiarini, Elena Roncaglione, Eleonora Carravieri, Emanuela Tenca & Maria T. Guasti. 2010. SLI children’s weakness in morphosyntax and pragmatics. Proceedings of the Tenth Tokyo Conference on Psycholinguistics, 57–75. Tokyo: Hituzi Syobo. Belletti, Adriana & Cornelia Hamann. 2004. On the L2/bilingual acquisition of French by two young children with different source languages. In Philippe Prévost & Johanne Paradis (eds.), The acquisition of French in different contexts, 147–174. Amsterdam/Philadelphia: Benjamins. Benmamoun, Elabbas, Silvina Montrul & Maria Polinsky. 2013a. Defining an “ideal” heritage speaker: Theoretical and methodological challenges. Reply to peer commentaries. Theoretical Linguistics 39 (3–4). 259–294. Benmamoun, Elabbas, Silvina Montrul & Maria Polinsky. 2013b. Heritage languages and their speakers: Opportunities and challenges for linguistics. Theoretical Linguistics 39 (3–4). 129–181. Bianchi, Giulia. 2012. Gender in Italian-German bilinguals: A comparison with German L2 learners of Italian. Bilingualism: Language and Cognition 16(3). 538–557. Bishop, Dorothy. 2003. Test for the Reception of Grammar – Version 2 (TROG-2). London: Pearson. Blom, Elma, Tessel Boerma & Jan De Jong. 2019. First Language Attrition and Developmental Language Disorder. In Monika S. Schmid & Barbara Köpke (eds.), Oxford Handbook of Language Attrition, 108–120. Oxford: Oxford University Press. Bortolini, Umberta, Barbara Arfé, Cristina M. Caselli, Luisa Degasperi, Patricia Deevy & Laurence B. Leonard. 2006. Clinical markers for specific language impairment in Italian: The contribution of clitics and non-word repetition. International Journal of Language and Communication Disorders 41. 695–712. Bortolini, Umberta, Cristina M. Caselli, Patricia Deevy & Laurence B. Leonard. 2002. Specific language impairment in Italian: first steps in the search of a clinical marker. International Journal of Language and Communication 37(2). 77–93. Bowles, Melissa A. 2011. Measuring implicit and explicit linguistic knowledge: What can heritage language learners contribute? Studies in Second Language Acquisition 33. 247–271. Brown, Roger. 1973. Development of the first language in the human species. American Psychologist 28(97). Cecchetto, Carlo, Alberto di Domenico, Maria Garraffa & Costanza Papagno. 2012. Comprendo. Batteria per la Comprensione di Frasi negli Adulti. Milano: Raffello Cortina Editore. Chamorro, Gloria & Antonella Sorace. 2019. The Interface Hypothesis as a framework for studying L1 attrition. In Monika S. Schmid & Barbara Köpke (eds.), The Oxford Handbook of Language Attrition.

90 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

Chamorro, Gloria, Antonella Sorace & Patrick Sturt. 2016. What is the source of L1 attrition? The effect of recent L1 re-exposure on Spanish speakers under L1 attrition. Bilingualism 19. 520–532. Clay, Marie M. 1971. Sentence repetition: elicited imitation of a controlled set of syntactic structures by four language groups. Monographs of the society for research in child development 36. 1–85. Cuza, Alejandro. 2010. The L1 attrition of the Spanish present tense. Hispania 93(2). 256–272. Cuza, Alejandro. 2013. Crosslinguistic influence at the syntax proper: interrogative subject-verb inversion in Spanish heritage speakers. International journal of bilingualism 17(1). 71–96. Cuza, Alejandro, Laura Miller, Rocio Pérez Tattam & Mariluz Ortiz Vergara. 2019. The case of the Spanish personal a. International Journal of Bilingualism 23(6).1333–1357. Cuza, Alejandro & Nelleke Strik. 2012. Patterns of morphosyntactic convergence and child L1 attrition: Evidence from subject-verb inversion in Spanish-English bilingual children. In Linguistic Symposium on Romance Languages (LSRL42). Southern Utah University, USA. D’Alessandro, Roberta. 2021. Syntactic Change in Contact: Romance. Annual Review of Linguistics 2021 7(1). 309–328. D’Alessandro, Roberta, David Natvig & Michael T. Putnam. 2021. Addressing challenges in formal research in moribund heritage languages: a path forward. Frontiers in psychology 12. Article 2589. Daskalaki, Evangelia, Elma Blom, Vasiliki Chondrogianni & Johanne Paradis. 2020. Effects of parental input quality in child heritage language acquisition. Journal of child language 47(4). 709–736. Dickson, Euan, Laura Manderson, Mateo Obregon & Maria Garraffa. 2021. Tracking Biliteracy Skills in Students Attending Gaelic Medium Education: Effects of Learning Experience on Overall Reading Skills. Languages 6(1). 55. Domínguez, Laura. 2009. Charting the route of bilingual development: contributions from heritage speakers’ early acquisition. International journal of bilingualism 13. 271–287. Domínguez, Laura. 2013. Understanding Interfaces: Second Language Acquisition and First Language Attrition of Subject Realization and Word Order Variation. Amsterdam: John Benjamins. Fattal, Iris, Naama Friedmann & Aviva Fattal-Valevski. 2011. The crucial role of thiamine in the development of syntax and lexical retrieval: a study of infantile thiamine deficiency. Brain 134. 1720–1739. Fenyvesi, Anna. 2005. Hungarian in the United States. In Anna Fenyvesi (ed.), Hungarian Language Contact outside Hungary: Studies on Hungarian as a Minority Language, 265–318. John Benjamins Ferrari, Ida. 2006. Acquisition of object clitics in Italian by two German/Italian bilingual children. Language Acquisition and Development. Proceedings of GALA 2005. 173–183. Flores, Cristina & Esther Rinke. 2020. The relevance of language-internal variation in predicting heritage language grammars. Bilingualism: Language and Cognition 23(1). 25–26. Friedmann, Naama. 1998. BAFLA: Friedmann’s battery for agrammatism. Tel Aviv: Tel Aviv University. Friedmann, Naama. 2001. Agrammatism and the psychological reality of the syntactic tree. Journal of Psycholinguistic research 30. 71–90. Friedmann, Naama. 2007. Young children and A-chains: the acquisition of Hebrew unaccusatives. Language acquisition 14. 377–422. Friedmann, Naama & Yosef Grodzinsky. 1997. Tense and agreement in agrammatic production: pruning the syntactic tree. Brain and language 56. 397–425. Friedmann, Naama & Hedva Lavi. 2006. On the order of acquisition of A-movement, Wh-movement and V-C movement. In Adriana Belletti, Elisa Bennati, Cristiano Chesi, Elisa Di Domenico & Ida Ferrari (eds.), Language Acquisition and Development, 211–217. Newcastle: Cambridge Scholars Press. Fishman, Joshua A. 2001. 300-plus years of heritage language education in the United States. In Joy Kreeft Peyton, Donald A. Ranard & Scott McGinnis (eds.), Heritage languages in America: Preserving

Chapter 3 Grammatical competence in adult heritage speakers 

 91

a national resource, 81–97. Washington, DC and McHenry, IL: Center for Applied Linguistics and Delta Systems. Granfeldt, Jonas & Suzanne Schlyter. 2004. Cliticisation in the acquisition of French L1 and L2. In Philippe Prévost & Johanne Paradis (eds.), The acquisition of French in different contexts, 333–369. Amsterdam/Philadelphia: John Benjamins. Guasti, Maria T. 1993/4. Verb syntax in Italian child grammar: Finite and non-finite verbs. Language Acquisition 3. 1–40. Guasti, Maria T., Silvia Palma, Elisabetta Genovese, Paolo Stagi, Gabriella Saladini & Fabrizio Arosio. 2016. The production of direct object clitics in pre-school–and primary school–aged children with specific language impairments. Clinical linguistics & phonetics 30(9). 663–678. Gürel Ayse. 2002. Linguistic Characteristics of Second Language Acquisition and First Language Attrition: Overt Versus Null Pronouns. Montreal, Canada: McGill University doctoral dissertation. Gürel Ayse. 2004. Selectivity in the L2-induced L1 attrition: a psycholinguistic account. Journal of neurolinguistics 17. 53–78. Håkansson, Gisela. 1995. Syntax and morphology in language attrition. A study of five bilingual expatriate Swedes. International Journal of Applied Linguistics 5. 153–171. Hamann, Cornelia. 2002. From syntax to discourse. Pronominal clitics, null subjects and infinitives in child language. Boston: Kluwer Academic Publishers. Hamann, Cornelia, Stéphanie Ohayon, Dubé Sébastien, Ulrich H. Frauenfelder, Luigi Rizzi, Michal Starke & Pascal Zesiger. 2003. Aspects of grammatical development in young French children with SLI. Developmental Science 6(2). 151–158. Hamann, Cornelia, Solveig Chilla, Lina Abed Ibrahim & István Fekete. 2020. Language assessment tools for Arabic-speaking heritage and refugee children in Germany. Applied Psycholinguistics 41(6). 1375–1414. Hulk, Aafke & Natascha Müller. 2000. Bilingual first language acquisition at the interface between syntax and pragmatics. Bilingualism: language and cognition 3(3). 227–244. Hyams, Nina & Jeanette Schaeffer. 2008. Clitic and auxiliary omissions in Italian children’s participle constructions. The Proceedings of the Inaugural Conference on Generative Approaches to Language Acquisition North America, 125–136. University of Connecticut Occasional Papers in Linguistics 4. Iverson, Michael B. 2012. On L1 Attrition and the Interface Hypothesis-A Case Study of Spanish Attrition in Contact with Brazilian Portuguese. Iowa: University of Iowa. Jakubowicz, Celia, Lea Nash, Catherine Rigaut & Christophe-Loic Gerard. 1998. Determiners and clitic pronouns in French-speaking children with SLI. Language acquisition 7(2–4). 113–160. Kayne, Richard. 1975. French syntax: The transformational cycle. Cambridge, Massachusetts: MIT Press. Kim, Ji-Hye, Silvina Montrul & James Yoon. 2009. Binding interpretation of anaphors by Korean heritage speakers. Language Acquisition 16 (1). 3–35. Kohnert, Kathryn J., Elizabeth Bates & Arturo Hernández. 1999. Balancing bilinguals: lexical-semantic production and cognitive processing in children learning Spanish and English. Journal of Speech, Language, and Hearing Research 42. 1400–1413. Köpke, Barbara. 2002. Activation thresholds and non-pathological L1 attrition. In: Franco Fabbro (ed.), Advances in the Neurolinguistics of Bilingualism, 119–142. Forum, Udine. Kupisch, Tanya. 2007. Determiners in bilingual German–Italian children: What they tell us about the relation between language influence and language dominance. Bilingualism: Language and Cognition 10(1). 57–78. Kupisch, Tanya & Jason Rothman. 2018. Terminology matters! Why difference is not incompleteness and how early child bilinguals are heritage speakers. International Journal of Bilingualism 22(5). 564–582.

92 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

Laleko, Oksana & Maria Polinsky. 2016. Between syntax and discourse: Topic and case marking in heritage speakers and L2 learners of Japanese and Korean. Linguistic Approaches to Bilingualism 6(4). 396–439. Leonard, Laurence B. & Marco Dispaldro. 2013. The effects of production demands on grammatical weakness in Specific Language Impairment: The case of clitic pronouns in Italian. Journal of Speech Language and Hearing Research 56. 1272–1286. Leonini, Chiara. 2006a. Object clitics and determiners in the acquisition of Italian as L1 and L2. Language Acquisition and Development. Proceedings of Gala 2005, 343–348. Newcastle upon Tyne: Cambridge Scholars Press. Leonini, Chiara. 2006b. The Acquisition of Object Clitics and Definite Articles. Siena: University of Siena doctoral dissertation. Leonini, Chiara & Adriana Belletti. 2004. Adult L2 acquisition of Italian clitic pronouns and subject inversion/VS structures. Language Acquisition and Development. Proceedings of GALA 2004, 293–304. Utrecht: LOT. Linck, Jared A., Judith F. Kroll & Gretchen Sunderman. 2009. Losing access to the native language while immersed in a second language. Psychological Science 20. 1507–1515. Lombardi, Linda & Mary C. Potter. 1992. The regeneration of syntax in short term memory. Journal of Memory and Language 31. 713–733. Lust, Barbara, Suzanne Flynn & Claire Foley. 1996. What children know about what they say: elicited imitation as a research method for assessing children’s syntax. In Dana McDaniel, Cecilia McKee, and Helen Smith Cairns (eds.), Methods for Assessing Children’s Syntax, 55–76. Cambridge, MA: MIT Press. Meir, Natalia, Joel Walters & Sharon Armon-Lotem. 2015. Disentangling SLI and bilingualism using sentence repetition tasks: the impact of L1 and L2 properties. International Journal of Bilingualism 20(4). 1–32 Moscati, Vincenzo & Roberta Tedeschi. 2009. The delay of Italian past participle agreement. Proceedings of the XXXIII Boston University Conference on Language Development, 379–390. Sommerville MA: Cascadilla Press. Marian, Viorica, Henrike K. Blumenfeld & Margarita Kaushanskaya. 2007. The Language Experience and Proficiency Questionnaire (LEAP-Q): Assessing language profiles in bilinguals and multilinguals. Journal of Speech, language, and hearing research 50. 940–967. Marinis, Theodoros & Sharon Armon-Lotem. 2015. Sentence repetition. In Suzanne Armon-Lotem, John H.A.L. de Jong & Natalia Meir (eds.), Assessing multilingual children: Disentangling bilingualism from language impairment, 95–122. Bristol: Multilingual Matters. Montrul, Silvina. 2004. Psycholinguistic evidence for split intransitivity in Spanish second language acquisition. Applied psycholinguistics 25(2). 239. Montrul, Silvina. 2008. Incomplete Acquisition in Bilingualism. Re-examining the Age Factor (Vol. 39). Amsterdam: John Benjamins. Montrul, Silvina. 2010. How similar are L2 learners and heritage speakers? Spanish clitics and word order. Applied psycholinguistics 31. 167–207. Montrul, Silvina. 2016. The acquisition of heritage languages. Cambridge: Cambridge University Press. Montrul, Silvina, Rebecca Foote & Silvia Perpiñán. 2008. Gender agreement in adult second language learners and Spanish heritage speakers: The effects of age and context of acquisition. Language learning 58. 503–553. Nagy, Naomi. 2016. Heritage languages as new dialects. The future of dialects: Selected papers from Methods in Dialectology XV. 15–34.

Chapter 3 Grammatical competence in adult heritage speakers 

 93

Nagy, Naomi. 2017. Documenting variation in (endangered) heritage languages: how and why? Language documentation & conservation special publication 13. 33–64. Nicoladis, Elena. 2012. Cross-linguistic influence in French–English bilingual children’s possessive constructions. Bilingualism: Language and cognition 15(2). 320–328. Paradis, Johanne. 2004. The relevance of specific language impairment in understanding the role of transfer in second language acquisition. Applied psycholinguistics 25. 67–82. Paradis, Johanne. 2010. The interface between bilingual development and specific language impairment. Applied psycholinguistics 31. 227–252. Peirce, Jonathan, Jeremy R. Gray, Sol Simpson, Michael MacAskill, Richard Höchenberger, Hiroyuki Sogo, Erik Kastman & Jonas Kristoffer Lindeløv. 2019. PsychoPy2: Experiments in behavior made easy. Behavior Research Methods 51. 195–203. Polinsky, Maria. 2008. Heritage language narratives. In Brinton, D., Kagan, O. & Bauckus, S. (eds.), Heritage language education. A new field emerging, 149–164. New York: Routledge. Polinsky, Maria 2011. Renalysis in adult heritage language: A case for attrition. Studies in second language acquisition 33, 305–328. Polinsky Maria. 2018. Heritage languages and their speakers. Cambridge: Cambridge University Press. Polinsky, Maria & Gregory Scontras. 2020. Understanding heritage languages. Bilingualism 23(1). 4–20. Potter, Mary. C. & Linda Lombardi. 1990. Regeneration in the short-term recall of sentences. Journal of memory and language 29. 633–654. Potter, Mary. C. & Linda Lombardi. 1998. Syntactic priming in immediate recall of sentences. Journal of Memory and Language 38. 265–282. Putnam, Michael T. & Joseph C. Salmons. 2013. Losing their (passive) voice: Syntactic neutralization in heritage German. Linguistic Approaches to Bilingualism 3(2). 233–252. R Core Team. 2020. R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing, Available online: https://www.R-project.org/. Riches, Nick G., Tom Loucas, Gillian Baird, Tony Charman & Emily Simonoff. 2010. Sentence repetition in adolescents with specific language impairments and autism: an investigation of complex syntax. International journal of language and communication disorders 45. 47–60. Rinke Eshter & Cristina Flores. 2014. Morphosyntactic knowledge of clitics by Portuguese heritage bilinguals. Bilingualism 17(4). 681–699. Rogers, Vivienne. 2009. Syntactic development in the second language acquisition of French by instructed English learners. Newcastle: Newcastle University doctoral dissertation. Romano, Francesco. 2020. Ultimate attainment in heritage language speakers: Syntactic and morphological knowledge of Italian accusative clitics. Applied psycholinguistics 41(2). 347–380. Romano, Francesco. 2021. L1 versus dominant language transfer effects in L2 and heritage speakers of Italian: A structural priming study. Applied linguistics 42(5). 945–969. Romano, Francesco. 2023 (this volume). Ultimate attainment of gender in heritage and L2 Italian. In Francesco Romano (ed.), Studies in Italian as a heritage language. Berlin: De Grutyer Mouton. Romano, F & Guijarro-Fuentes, P. (Under review). Task effects and the yes-bias in heritage language acquisition. Under consideration at the International Journal of Bilingualism and Bilingual Education. Rothman, Jason. 2009. Understanding the nature and outcomes of early bilingualism: Romance languages as heritage languages. International Journal of Bilingualism 13(2). 155–163. Schaeffer, Jeannette. 2000. The acquisition of direct object scrambling and clitic placement. Syntax and Semantics. Schmid, Monika S. 2011. Language attrition. Cambridge University Press.

94 

 Giuditta Smith, Roberta Spelorzi, Antonella Sorace, Maria Garraffa

Serratrice, Ludovica, Antonella Sorace & Sandra Paoli. 2004. Subjects and objects in Italian-English bilingual and monolingual acquisition. Bilingualism: Language and Cognition 7. 183–206. Smith, Giuditta. 2021. Patterns in clitic pronouns: Assessment of clitics in Italian typical and nontypical populations. Rovereto: Università di Trento doctoral dissertation. Sorace, Antonella. 2005. Selective optionality in language development. Cornips, Leonie &Corrigan, Karen P. (eds.), Syntax and Variation: Reconciling the Biological and the Social, 55–80. Amsterdam, Netherlands: John Benjamins Publishing Company. Sorace, Antonella. 2011. Pinning down the concept of “interface” in bilingualism. Linguistic approaches to bilingualism 1(1). 1–33. Sorace, Antonella & Francesca Filiaci. 2006. Anaphora resolution in near-native speakers of Italian. Second language research 22(3). 339–368. Spelorzi, Roberta, Giuditta Smith, Antonella Sorace & Maria Garraffa. 2022. Generational attrition: first language attriters and heritage speakers on production of Italian clitic pronouns. Paper presented at the Incontro di Grammatica Generativa 47 (IGG 47), Catania, 24–26 February, 2022. Sukenik, Nufar & Naama Friedmann. 2018. ASD is not DLI: Individuals with autism and individuals with syntactic DLI show similar performance level in syntactic tasks, but different error patterns. Frontiers in psychology 9. 279. Tsimpli, Ianthi. 2007. First language attrition from a minimalist perspective. In Köpke, Barbara, Schmid, Monika S, Keijzer, Merel & Dostert, Susan (eds.), Language attrition: Theoretical perspectives. 83–98, Amsterdam, Netherlands: John Benjamins Publishing Company. Tsimpli, Ianthi, Antonella Sorace, Caroline Heycock & Francesca Filiaci. 2004. First language attrition and syntactic subjects: A study of Greek and Italian near-native speakers of English. International Journal of Bilingualism 8(3). 257–277. Tuller, Laurice, Hélène Delage, Cécile Monjauze, Anne-Gaëlle Piller & Marie-Anne Barthez. 2011. Clitic pronoun production as a measure of atypical language development in French. Lingua 121(3). 423–441. Tuller, Laurice, Cornelia Hamann, Solveig Chilla, Sandrine Ferré, Eleonore Morin, Philippe Prévost, Phillipe, Christope dos Santos, Lina Abed Ibrahim & Racha Zebib. 2018. Identifying language impairment in bilingual children in France and in Germany. International Journal Language Communication Disorders 53. 888–904. Yip, Virginia, and Stephen Matthews. 2000. Syntactic transfer in a Cantonese–English bilingual child. Bilingualism: Language and cognition 3(3). 193–208. Vender, Maria, Maria T. Guasti, Maria Garraffa & Antonella Sorace. 2016. How early L2 children perform on Italian clinical markers of SLI: a study of clitic production and nonword repetition. Clinical linguistics and phonetics 30 (2). 150–169. White, Lydia. 1996. Clitics in L2 French. In Harald Clahsen (ed.), Generative perspectives on language acquisition, 335–368. Amsterdam: Benjamins.

Francesco Romano

Chapter 4  Ultimate attainment of gender in heritage and L2 Italian Abstract: Studies of gender in various L2s have proven this area to be particularly difficult to master, even for highly proficient L2 speakers (Franceschina, 2005; McCarthy, 2009; Grüter et al., 2012, inter alia). Although disagreement exists as to why this might be, L2 learners characteristically: (1) display low accuracy levels for gender agreement and/or assignment compared to L1 speakers; (2) adopt a default form, typically the masculine form overextending to feminine contexts in dual gender systems like Spanish. Two views vie to explain non-target-likeness, representational versus computational. Representational accounts such as the failed functional feature hypothesis (FFFH), for which age of onset and L1 transfer play a deterministic role on the acquisition of uninterpretable features, compete with computational accounts such as the missing surface inflection hypothesis (MSIH), for which difficulty in retrieval of appropriate spell-out forms under task pressure accounts for non-target-likeness and is inconsequential to features’ underlying representation. Heritage research, however, allows researchers to test age of onset in the acquisition of gender given heritage language speakers are often raised bilingually from very early ages. Despite results from heritage studies showing a remarkable convergence to L2 speakers (Rodina et al. 2020; Bianchi, 2012; Polinsky, 2008), the FFFH and MSIH have not yet been sufficiently considered with HL speakers. The present chapter reports a study aimed at further testing the plausibility of the FFFH and MSIH for heritage language populations and better explaining variability in knowledge of grammatical gender in heritage grammars. The ultimate attainment of Italian gender was scrutinized by comparing highly-proficient Swedish-dominant L2 speakers to highly-proficient Swedish-dominant heritage speakers of Italian and L1 speakers of Italian in Italy. An oral structural priming task and a timed GJT elicited production, judgment, and response time data on knowledge of masculine and feminine gender assignment between nouns and agreeing accusative clitics. Consistent with previous L2 and heritage research, results show the heritage and L2 speakers are more alike than the monolinguals in terms of accuracy, suggesting age of onset is not implicated in ultimate attainment as posited by the FFFH. Additionally, the heritage group was significantly faster in responding correctly to feminine items, suggesting masculine forms to be particularly vulnerable during the processing of gender. This finding is conhttps://doi.org/10.1515/9783110759587-005

96 

 Francesco Romano

sistent with processing accounts such as the MSIH which posit masculine forms to act as defaults. I  argue that, despite most results being consistent with the MSIH and the presence of gender in the L1 playing a facilitative role in ultimate attainment of Italian gender, recourse to grammar external factors need complement current theories proposed to explain non-target-likeness in heritage language knowledge. Keywords: Gender, FFFH, MSIH, age of onset, transfer

1 Introduction Inflectional morphology is fairly uncontroversially held to be a vulnerable domain in SLA and heritage language acquisition (White, 2003; Montrul, 2016; Polinsky, 2018). One area that has been particularly well-researched, in this respect, is the acquisition of gender (for Italian, Chini, 1995, 1998; Oliphant, 1998; Bianchi, 2012; for French, Dewaele & Veronique, 2001; Granfeldt, 2005; Renaud, 2009; for Spanish, Franceschina, 2001, 2005; Grüter, Lew-Williams & Fernald, 2012; Hawkins & Franceschina, 2004; McCarthy, 2009; Montrul et al., 2008; White, Valenzuela, Kozlowska & Leung, 2004; for Dutch, Sabourin, Stowe  & de Haan, 2006; for German, Matteini, 2010; Spinner & Juffs, 2008; inter alia). Studies of gender which have focused on both gender assignment and gender agreement phenomena have found consistent divergence between second language speakers (L2Ss) and native speakers (L1s), especially in language production where non-target-like forms are used even by L2 learners at very advanced levels who have received considerable instruction (Franceschina, 2005; McCarthy, 2009; Grüter et al., 2012, inter alia). The consistent variability in knowledge of gender observed in L2 learners is also characterized by the use of one particular form of gender which overextends to other contexts (e.g. masculine for feminine), acting as a default in both production and comprehension (McCarthy, 2009; Grüter et al., 2012, Montrul et al. 2008). These findings also resonate widely in the heritage language literature (Rodina, Kupisch, Meir, Mitrofanova, Urek, and Westergaard, 2020; Håkansson, 1995; Montrul et al. 2008; Hur, Lopez-Otero, and Sanchez, 2020, inter alia). The ultimate attainment of gender is, thus, characterized by divergence from L1s in both agreement and assignment, non-target-like use in production, and recourse to a default in production and comprehension. For the purpose of the present study, ultimate attainment is defined as ‘the state of knowledge actually attained at the end or steady state of grammatical development’ (Lardiere, 2007, p.5). Two competing views have been proposed to explain variability in L2 inflectional morphology, the computational and representational. Within these camps,

Chapter 4 Ultimate attainment of gender in heritage and L2 Italian 

 97

two hypotheses have been especially influential, the Failed Functional Feature Hypothesis (FFFH) (Hawkins and Chan, 1997) and the Missing Surface Inflection Hypothesis (MSIH) (Prévost and White, 2000). According to the FFFH, a representational view, age of onset and the availability of an L2 feature in the L1 play a deterministic role on the acquisition of gender at an abstract level. By contrast, a computational view such as the MSIH posits that task pressure and non-target-like lexical retrieval are responsible for native-L2 divergence without any consequence to knowledge of abstract features associated to gender. These hypotheses are all the more appealing for explaining the ultimate attainment of gender given that, as explained above, several findings of variability in L2 acquisition are shared with heritage language acquisition. Advances in heritage language (HL) research have, moreover, crucially allowed researchers to look further into age of onset as a key factor in the acquisition of gender, as claimed by the FFFH, given that HL speakers (HLSs) are often raised bilingually from very early ages. This has been done by extemporaneously comparing knowledge of inflectional morphology, including but not limited to gender, in HL, L2, and L1 groups (Alarcon, 2011; Au et al., 2002; Håkansson, 1995; Montrul et al. 2008). Therefore, despite results obtained from HLSs showing a remarkable overlap with L2Ss, the FFFH and MSIH have yet not been sufficiently put to the test with the former population (Montrul, 2011, Montrul et al. 2008). The goal of the present study is two-fold. For one, it aims at explaining the characteristic variability in knowledge of gender by L2Ss in comparison to L1s by specifically examining factors such as age of onset, L1 transfer, task pressure, and lexical retrieval in relation to the FFFH and MSIH. Secondly, it extends the strand of conversation revolving around computational versus representational accounts to HLSs who represent an ideal testing ground with respect to age of onset. In this study, the ultimate attainment of Italian gender was scrutinized by comparing highly-proficient adult Swedish-dominant L2Ss to proficiency-matched adult Swedish-dominant HLSs of Italian as well as L1s. Given that Swedish and Italian both have binary, though not identical, gender systems, the languages selected are particularly befitting for testing the FFFH’s claim that the presence of an L2 feature in the L1 plays a facilitative role in L2 acquisition.

2 The L2 acquisition of gender Gender is one of the most extensively studied properties is L2 acquisition. Ultimate attainment and acquisition of the gender system of Romance and Germanic languages by adult L2Ss has been investigated for Italian (e.g., Chini, 1995, 1998; Oliph-

98 

 Francesco Romano

ant, 1998), French (Dewaele  & Veronique, 2001; Granfeldt, 2005; Renaud, 2009), Spanish (Franceschina, 2001, 2005; Grüter, Lew-Williams & Fernald, 2012; Hawkins & Franceschina, 2004; McCarthy 2007, 2009; Montrul et al., 2008; White, Valenzuela, Kozlowska & Leung, 2004), Dutch (Sabourin, Stowe & de Haan, 2006), and German (Matteini, 2010; Spinner & Juffs, 2008). Within this wide literature, two aspects of gender have received special focus, gender assignment and gender agreement. Consider (1) and (2): (1) La sedia rossa the.SG.FEM chair[SG.FEM] red.SG.FEM ‘The red chair’ (2) Il palloncino bianco the.SG.MASC balloon[SG.MASC} white.SG.MASC ‘The white balloon’ In Romance languages such as Italian, gender is marked in two distinct ways, lexically and syntactically. On the one hand, nouns such as sedia ‘chair’ and palloncino ‘balloon’ are lexically assigned masculine (MASC) or feminine (FEM) gender already in the lexicon. Native speakers learn that each word must be assigned one and only one of two genders to form part of a given word’s lemma. In (1), thus, native speakers know that sedia and palloncino are respectively FEM and MASC words. On the other, such knowledge enables them to compute gender agreement syntactically in a correct manner, namely by assigning the appropriate lexical gender features to other constituents that they agree with such as articles and adjectives, la and rossa in (1) or il and bianco in (2). However, nouns can also agree with referential items such as direct pronouns which are dealt with in greater detail in this chapter. By now, robust evidence exists of variability in L2 grammars both for gender agreement and assignment. In a study of L2 Spanish, for instance, McCarthy (2009) showed that intermediate and advanced adult English learners of Spanish display systematic and highly predictable variability in both the production and comprehension of gender agreement. An elicited imitation test was used to examine agreement between a full NP, a determiner, and an accusative clitic. L2 participants to the study were intended to produce sentences such as Tiene una manzana. Está comiéndola where the NP.SG.FEM manzana ‘apple’ agrees with the determiner una ‘a’ with features SG and FEM as well as the pronoun la ‘it’ cliticised to the verb comer ‘to eat’. McCarthy found that both the intermediate and advanced group accuracy levels were significantly lower than a control group of native Spanish speakers. A closer look at the errors committed revealed correct clitic and adjective agreement

Chapter 4 Ultimate attainment of gender in heritage and L2 Italian 

 99

was far greater with masculine than feminine gender. This asymmetry was due to a consistent tendency to overextend masculine clitics to feminine contexts and not vice versa (i.e. masculine lo to refer to feminine la in la ventana ‘the window’). This tendency was then also attested in a picture selection task where participants were asked questions like: Paco quiere llevar algunas cosas que acaba de comprar pero no encuentra nada. Paco dice: ‘Acabo de comprarlo: ¿dónde está?’ Paco wants to bring some things that he just bought, but he can’t find anything. Paco says, ‘I just bought it-SG.MASC: Where is it?’ (p.472)

Participants were expected to select one correct picture out of a possible three, specifically the one depicting a SG.MASC NP to agree with the clitic lo. The intermediate group of L2Ss was significantly less accurate than native controls and were worse in the MASC gender condition than the FEM, consistent with the result found for production. The advanced group did not differ from native controls but also scored higher in the FEM condition, implicating variability in gender assignment does not extend to comprehension at advanced levels. All in all, this study then shows, first of all, that ultimate attainment (i.e. the performance of near-natives) of gender agreement is problematic for production but not comprehension. A key difference between the two tasks was the task pressure imposed on retrieval of the appropriate gender form by the production task which may be responsible for the native vs. near-native differences and the variability observed. Moreover, this study suggests a bias in L2 Spanish grammars for masculine defaults which tend to be overextended to female contexts in both comprehension and production (see also Shin, Rodríguez, Armijo & Perara-Lunde, 2019 and Goebel-Mahrle & Shin, 2020 for studies of children). In a series of L2 Spanish experiments, Grüter et al. (2012) showed that advanced adult English learners of Spanish display similar variability also in the production, comprehension, and processing of gender assignment. Near-native English speakers of Spanish were tested alongside native speakers of different varieties of Spanish on similar elicited production and comprehension tasks as McCarthy (2007, 2009), as well as a looking-while-listening task intended to measure processing. The latter task comprised familiar and novel (i.e. nonce) words in order to quantify effects of exposure to familiar and unfamiliar vocabulary. Because native speakers have greater exposure than L2Ss to familiar words, the use of novel words nullifies any advantages for the former group. In the first task, the near-native speakers were significantly less accurate than the natives, scoring 80 vs. 98.7% respectively. 17.2 of the 20% error rate, however, were gender assignment rather than agreement errors which consisted of determiner + adjective responses mismatching a refer-

100 

 Francesco Romano

ring full NP. When shown a picture of two butterflies of different colors and asked ¿Cuál mariposa prefieres? ‘Which butterfly do you like better?’, the near-natives on average replied with responses such as el rojo ‘the red’ where the article and adjective are both incorrectly MASC against the FEM gender of mariposa and coded thus as an error of gender assignment. This result confirms, as in McCarthy, that production can be problematic and that masculine forms act as defaults in these interlanguages. Unlike McCarthy, however, Grüter et al. show gender assignment to be difficult to attain. Much like McCarthy, the picture-selection task, nevertheless, yielded ceiling performance for both the L2 and control groups, suggesting nativelike comprehension of gender is attainable. Lastly, in the looking-while-listening task, it was found that overall the two groups did not differ significantly from each other in the novel word condition. Because the novel words were, by definition, lexical items neither group had been previously exposed to in comparison to the familiar words, results from the processing task imply that native-like processing of gender assignment is indeed attainable by near-native speakers but sensitive to exposure effects.

3 Variability in morphological knowledge Two prominent views are equipped to account for the variability observed in the production of morphosyntax by advanced L2Ss. The Representational view claims that when a particular L2 morphosyntactic feature is not available in the L1, the feature will no longer be accessible via UG (Hawkins & Chan, 1997; Hawkins and Franceschina, 2004; Hawkins, 2003; Hawkins and Liszka, 2003; Hawkins and Hattori, 2006; Tsimpli and Dimitrakopoulou, 2007; Tsimpli and Mastropavlou, 2008). Within the representational camp, views further differ on the basis of whether maturational effects and feature interpretability also play a part in the unacquirability of a feature. To name one, the Failed Functional Feature Hypothesis (Hawkins and Chan, 1997) claims that an L2 feature is fully acquirable if it is instantiated in the L1 and acquired before onset of the critical period. In consequence, a failed functional feature results in variability which in experimental terms amounts to accuracy levels significantly lower than native speakers and use of non-target like forms such as misagreement and omission in production. These deviant forms are assumed to be the result of a feature being either underspecified or lacking from L2 abstract representations. In addition, according to the FFFH, variability is also expected to manifest itself under different test conditions insofar as a missing or underspecified functional feature will affect the grammar regardless of elicitation method.

Chapter 4 Ultimate attainment of gender in heritage and L2 Italian 

 101

A second view, referred to as computational, maintains that ultimate attainment of features not present in the L1 is indeed possible (Haznedar & Schwartz, 1997; Prèvost  & White, 2000; Gess and Herschensohn, 2001; Lardiere, 2008;). Syntactic representations are intact and ultimate attainment of features not instantiated in the L1 and acquired after the critical period is not inevitable. Morphological variability in L2 production is the result of performance-related factors tied to, for instance, pressure on lexical retrieval during production (e.g. the Missing Surface Inflection Hypothesis (MSIH), Prèvost  & White, 2000) or the extent to which the features composing the L2 morpheme subject to variability in production need be re-arranged from L1 to L2 (the Feature Reassembly Hypothesis, Lardiere, 2008). In consequence, a computational view such as the MSIH predicts variability and non-target like forms are to be found only in production or other tasks where lexical retrieval is similarly taxed by time-pressure. A secondary prediction is that default gender be present across tasks as the non-target-like forms retrieved under pressure should be the same regardless of task type for the problem to be computational (McCarthy, 2007, 2009). In the method section, I will formulate a series of predictions along the lines above to put the two views to the test with the ultimate attainment of gender in L2 and heritage Italian speakers.

4 Heritage language acquisition of gender The acquisition of gender has been studied far less in heritage than L2 acquisition. Rodina, et al. (2020) recently reported the production of gender in Russian by 5 groups of English, Latvian, German, Norwegian, and Hebrew-dominant children aged 3–10. The goal of their study was, among others, to examine the role of cross-linguistic influence between the dominant and heritage language with respect to gender agreement. Three types of gender systems were distinguished: Type I languages with transparent gender-marking such as Italian and Spanish; Type II languages with semi-transparent gender-marking such as Russian, Latvian, and Hebrew; and Type III languages with opaque gender-marking such as Norwegian and Swedish. Their findings include no cross-linguistic influence and the use of masculine gender as default across the gender conditions tested. Several studies of gender have compared adult heritage speakers with L2Ss (Alarcon, 2011; Au et al., 2002; ; Håkansson, 1995; Montrul et al. 2008; Montrul, Davidson, De La Fuente & Foote, 2014; Montrul, De La Fuente, Davidson & Foote, 2013; Romano, 2020). These studies have mostly shown similar degrees of variability in the production of gender between the two populations. Montrul et al. (2008) who compared HLSs, L2Ss, and monolingual speakers of Spanish on knowl-

102 

 Francesco Romano

edge of gender within the DP found that gender agreement, as well as assignment, were affected in both groups to a comparable extent. Moreover, this study was the first to extend testing of the plausibility of representational versus computational accounts to explain HL knowledge of gender agreement and assignment. Crucially, this study found that (1) variability in HLSs resembles that of L2Ss more than monolingual controls and (2) HLSs were more accurate with gender agreement in an oral than written production task compared to L2Ss where the opposite pattern was attested. On the basis of (1), Montrul et al. (2008) concluded that computational accounts better explained this part of the HL data. However, a number of explanations were adduced to account for (2), namely that HLSs perform better on spoken production than L2Ss while the latter perform better than HLSs on written production tasks. On one account, this result is compatible with a representational view, provided that oral tasks are taken as evidence of implicit linguistic knowledge (p. 541). Au et al. (2002) also found that L2 and heritage speakers perform similarly in marking gender agreement within the Spanish DP. Thirdly, in a study of heritage Swedish, Håkansson (1995) observed that heritage speakers perform similarly to L2 learners in the production of DP-internal and DP-external gender agreement. In particular, she observed that heritage speakers had a reduced system of agreement suffixes where the unmarked form was overused and functioned as a default. Finally, Alarcon (2011) found that advanced L2 and HLSs of Spanish show similar knowledge of gender in comprehension and oral production tests. Errors in oral production were claimed to be due to difficulties in the surface manifestations of the abstract features of gender, in line with the Feature Reassembly Hypothesis. Studies focused purely on heritage interlanguage report similar findings with respect to variability. Polinsky (2008) provided evidence that the Russian gender system of American-Russian heritage speakers undergoes reanalysis as a result of possible changes in the assignment rules of the heritage language. Errors in the production of gender agreement, which largely resembles the errors commited by L1 children, were largely attested. Anderson (1999) found profuse gender agreement errors in the Spanish of two Spanish–English bilinguals after two years of residence in an Anglophone country. Moreover, gender agreement errors in the DP domain were also found in the production of heritage speakers of Spanish investigated by Lipski (1993). Gender marking has been investigated in heritage languages also in terms of lexical frequency effects even though these have been shown to be task-dependent. In other words, variability in the production of gender-marking is subject to previous knowledge of the meaning of the gender-marked nouns per sé but such effect varies by task. Hur et al. (2020) found that masculine gender is overextended to feminine contexts but that knowledge of the vocabulary to be gender-marked leads

Chapter 4 Ultimate attainment of gender in heritage and L2 Italian 

 103

to higher accuracy in elicited production. In cases where feminine gender was overextended to masculine contexts, no productive vocabulary knowledge effects were found, at least for a forced-choice task. Errors and non-target-like gender agreement and assignment are attested in 2nd generation HLSs of Italian in Australia and the USA by Bettoni (1991) and Scaglione (2000) respectively. More recently, Bianchi (2012) investigated ultimate attainment of gender assignment and agreement in three groups: (1) HLSs of Italian dominant in German, (2) HLSs of Italian dominant in Italian, and (3) L2Ss of Italian with L1 German. One of the goals of Bianchi’s study was to test the hypothesis that non-target-like use of gender in production and the associated mastery of uninterpretable features result from impoverished language use and input. In her study, Bianchi employed an elicited production task and an AJT controlling for the word classes to which the gender-marked words belonged. Her results show that groups 1 and 3 were more accurate in gender agreement than assignment while they were significantly less accurate than group 2 for both agreement and assignment. These findings are interpreted to suggest that larger amounts of input, which characterized the Italian of group 2, lead to more complete ultimate attainment of gender, since group 2 outperformed both group 1 and 3 for both assignment and agreement. Bianchi also interpreted her results to support the MSIH on grounds that agreement errors occurred mostly in the production task. Recall that one prediction made by the MSIH is that variability is more manifest in tasks that induce higher pressure on retrieval of the appropriate lexical spell-out for morphosyntactic features such as production tasks. However, the results show a trend in task effects which is not discussed by the author: although it is true that gender agreement scores tended to be higher in the AJT than the production task for 2 of the 3 groups, consistent with the MSIH, all mean accuracy scores for assignment tended to be higher in the production rather than the AJT task for all 3 groups (similarly to Montrul et al. 2008), contrary to the MSIH. The later finding, thus, speaks against task pressure as responsible for variability in, at least, gender assignment and supports representational accounts such as the FFFH as discussed previously for Montrul et al. (2008). Furthermore, a caveat of this study is that target-like knowledge of gender could not be established without a control group of monolingual speakers. Therefore, it is unclear whether the tasks employed elicited target-like rates of accuracy since no benchmark as such was determined. In this study, we address this issue by including a group of monolingual controls alongside HLSs and L2Ss. Another important contribution of the present study is the possibility of positive transfer from a language with gender to another, namely from Swedish to Italian. As will be seen in the next section, although the pronominal systems of the two languages are not entirely identical, their gender systems are both binary which may play a facilitative role in ultimate attainment. This aspect is particularly relevant to the field of

104 

 Francesco Romano

heritage language bilingualism as studying Italian allows for meaningful comparisons to heritage Spanish and testing of key findings for this language. In particular, in this section it was seen that masculine forms act as defaults in this language (Hur et al., 2020 inter alia).

5 Gender in Italian and Swedish 5.1 Gender in Italian In Italian, gender assignment occurs at the lexical level where either MASC or FEM gender is assigned as previously shown in (1)-(2). Assignment follows both semantic and morphophonological rules. Semantically, animate nouns mark gender reflecting their biological sex (cane “the he-dog”, cagna “the she-dog”) while inanimate nouns are assigned gender by semantic class. Names of fruits, thus, are typically FEM if they represent countable samples (mela {apple, FEM}) but MASC if denoting the uncountable tree to which the fruit belongs (melo {apple, MASC}) or MASC if the fruit’s lexeme is a loan word (mango {apple, MASC}). On the other hand, morphophonological rules determine a noun’s gender based on word endings following the set of declension classes outlined in Figure 1.

Figure 1: Gender declension classes in Italian. Note. From Chini (1995, p.81).

Italian nouns ending in -o which form class I are usually MASC (libro {book, MASC}) while those ending in -a which form class II are FEM (carta {paper, FEM}). However, by exception to this rule, several words ending in -o denoting body parts are MASC

Chapter 4 Ultimate attainment of gender in heritage and L2 Italian 

 105

when singular but FEM when plural, forming class VI (dito {finger, SG, MASC} → dita {finger, PL, FEM}). Similarly, several nouns ending in -a of Greek origin are MASC (problema {problem, MASC}; pianeta {planet, MASC}), forming class V. Finally nouns ending in -e are the least predictable based on morphophonological rules. The subdivision in Figure 1 is based not only on the morphophonological composition of nouns but also their frequency (Chini, 1998). Notably, nouns from class I and II which figure 71.5% of the time are more frequent than those in class III which figure 20.6%, class IV 5.4%, class V 1.2%, and VI and VII >.2%. Because the last four classes mainly constitute exceptions, the present study will focus on the three most frequent classes. According to Rodina et al. (2020), because of the fairly transparent morphophonological cues of its declension classes, Italian may be considered to have a fairly transparent gender-marking system. This view is supported by the observation that, in the language learning process, canonical word-endings for classes I and II tend to provide unambiguous cues for a particular gender as seen in Figure 1. Moreover, other gender-marked elements such as articles, determiners, and adjectives are either phonologically similar or identical to the canonical endings as shown in (3) where all constituents taking part in gender agreement are marked with FEM -a: (3) La mia sedia rossa the.ART.SG.FEM my.DET.SG.FEM chair[SG.FEM] red.ADJ.SG.FEM ‘my red chair’ Thirdly, Italian NPs are typically article + N structures which amounts to bare nominals being rare, providing unambiguous cues to the gender of a given noun in the learning process. Gender agreement between an Italian noun and another constituent in the sentence such as a clitic can be described by several approaches to syntactic features (Adger, 2002, p.41; Roberts, 2009, p.66; Carstens, 2000; Abney, 1987). These approaches differ as to the feature composition of the noun and agreeing clitic (Roberts, 2009), whether features are checked and matched (Adger, 2002) or shared (Pesetsky & Torrego, 2007), and whether gender is a functional projection in the DP (Abney, 1987). Following Chomsky (1995), I assume gender is an interpretable lexically-assigned phi (φ) feature (F) on nouns but uninterpretable feature (uF) on clitics which needs to be checked, valued and deleted before spell out. Valuation, checking, and deletion are made possible through the operation Agree between a Goal and Probe as in (4):

106  (4)

 Francesco Romano

agreement target (probe)

... ...

... ...

agreement controller (goal)

Adopting Pesetsky  & Torrego (2004), an unvalued uF of the clitic probe scans its c-command domain for a valued gender F on a goal noun with which to agree. Thus in a clitic-left dislocation (CLLD) structure such as (5), the goal noun pesci with valued gender values the probe clitic li for the same feature. (5) I pesci Pietro li the fish N.ACC.PL.MASC (goal) Pietro them.cl.ACC.[Gen:MASC] (probe) cucina all’aperto cooks outdoors ‘The fish, Pietro cooks them outdoors’ Given clitics adjoin to their verbal host, an analysis is necessary to distinguish Agree between nouns and accusative clitics as in Italian and nouns and accusative-marked full DPs in the other language of interest, namely Swedish. Following Roberts (2009), I consider clitics to be φmin/max, acategorial syntactic objects consisting only of the inflectional part of the structure of a pronoun Dmin/max. Simplifying the analysis slightly, the Agree relations in (6) are then able to account for differences in clitic-noun agreement in Italian versus DP-noun agreement in Swedish: (6) a. Trigger for Agree (Italian) φmin/max [Gen:__] b. Outcome of Agree φmin/max [Gen:MASC/FEM]) (7) a. Trigger for Agree (Swedish) DPmin/max [D, Gen:__] b. Outcome of Agree DPmin/max [D, Gen:U/NU]

N[Gen:MASC/FEM] N[Gen:MASC/FEM]

N[Gen:U/NU] N[Gen:U/NU]

In (6) and (7), N values the Gen feature of the clitic and the DP respectively, as indicated by the boldface, but φmin/max and DPmin/max differ lexically as the former lacks the D feature of the latter. The key aspect underlined by Roberts’ analy-

Chapter 4 Ultimate attainment of gender in heritage and L2 Italian 

 107

sis is, that despite both Swedish and Italian representing Gen, a non-uniformity between the DP/clitic features in relation to gender agreement of the two language is present.

5.2 Gender in Swedish Gender assignment in Swedish occurs at the lexical level where either uter (U) or neuter (NU) gender is assigned (see (7)). Approximately 75% of all nouns in Swedish are uter which holds for both oral and written, formal and informal discourse (Nyqvist & Lahtinen, 2021). Assignment follows both semantic and morphophonological rules but the relationship between semantics and gender choice is far less predictable in Swedish than Italian. While animate nouns, for instance, are utermarked –neuter barn ‘child’ as a notable exception– the gender of inanimate nouns can hardly ever be predicted on inanimacy. Morphophonological rules such as a word ending in -ing are highly productive in that all such words are uter but for the most part word form-to-gender mapping is highly unpredictable, requiring that gender be learned by rote (Nyqvist & Lahtinen, 2021). Nor can gender-marking be predicted based on shifting from a semantic class to another, as previously seen for Italian: shifts in the meaning of the same word form by gender are highly unsystematic and rare (e.g. en plan ‘an open space’ is U while ett plan ‘a floor” is NU). The morphophonological rules outlined in Table 1 capture the 5 main declension classes of nouns in Swedish: Table 1: Gender declension classes in Swedish. Class

Final sounds in pl. Gender

Example SG/PL

Translation

I II III

-or -ar -(e)r

IV V

-(e)n bare

en lampa/flera lampor en säng/flera sängar en madrass/flera madrasser ett land/flera länder ett täcke/flera täcken en läkare/flera läkare ett lakan/flera lakan

A lamp/several lamps A bed/several beds A mattress/several mattresses A country/several countries A duvet/ several duvets A doctor/several doctors A sheet/several sheets

U U U NU NU U NU

Note. Final sounds in singular always vary. From Levy-Scherrer and Lindemalm (2014, p.231) and Stensson (2013).

Unlike Italian where a noun’s declension class is determined in most cases by its ending in the singular form, Swedish declension classes are for the most part distinguishable by a noun’s plural form. Moreover, Swedish nouns never change gender

108 

 Francesco Romano

when pluralized as in Italian declension class VI. Much like Italian, however, MASC/ FEM gender and U/NU gender appropriate their own classes, respectively I and II in Figure 1 and I and IV in Table 1. Moreover, two classes allow nouns to be assigned either gender, III and IV in Figure 1 and III and V in Table 1. According to Rodina et al. (2020), Swedish is a language with an opaque gender system. Much like Norwegian, Swedish gender is hardly predictable based on semantic or morphophonological cues. Although both Swedish and Italian present neatly defined declensions classes, the only reliable morphophonological cues in Swedish are inflectional markings in the plural. Even though these cues apply consistently, the gender of a noun cannot always be unequivocally deduced as classes III and V can be of either gender. Thus, HL and L2Ss of Italian who are dominant in Swedish can be thought of as raised to face greater challenges in the learning of gender in Swedish than Italian.

6 The study The following research questions were addressed: 1. Which view can best explain variability in knowledge of L2 gender, the computational or representational? 2. To what extent can these two views explain knowledge in HLSs with respect to gender? In the process of addressing the aforementioned questions, two competing accounts within the computational and representational camps are put to the test, the FFFH and the MSIH. The FFFH makes three predictions: (1) accuracy levels for L2Ss but not HLSs are expected to be significantly lower than L1s owing to HL first exposure prior to the critical period; (2) variability is expected to manifest itself in the form of misagreement and omission in production; (3) variability is expected under different test conditions. By contrast, the MSIH predicts that: (1) HL and L2 groups may not differ in terms of accuracy as age of onset plays no deterministic role for computational accounts; (2) a default form for gender such as the MASC form in L2/ heritage Spanish will be the same across different types of tasks; (3) variability is to be found more so in tasks that impose greater time-pressure such as production. With regards to prediction (2), under the MSIH, one particular form is expected to overtextend over another as learners tend to privilege a specific non-target-like form during lexical retrieval. On the basis of previous research (e.g. Hur et al., 2020) and the typological similarity between Italian and Spanish, this form can be expected to be masculine.

Chapter 4 Ultimate attainment of gender in heritage and L2 Italian 

 109

7 Method 7.1 Participants HLSs (n = 14), L2Ss (n = 13), and L1s (n = 18) of Italian took part in the study. The bilingual groups were similar in proficiency as established by a C–test previously trialed with German and Swedish HLSs of Italian (Kupisch et al. 2012; Author 2020, 2021). Mean scores for the two groups (Table 2) were not significantly different (two tailed, t (21) = -0.54, p = .43), though both groups were significantly less proficient than the monolinguals (L2 vs. L1: two tailed, t (19) = –3.64, p < .01; HL vs. L1: two tailed, t (14) = –4.02, p < .01). Data collected from a background questionnaire documents that the whole of the HLSs and L2Ss were born and raised in Sweden, were dominant in Swedish, and self–assessed themselves as native speakers of Swedish. Most had a degree in Italian, were educated to university-level, and had travelled to Italy periodically during their lifetime. Half of the HL sample received some instruction in the mother tongue as children and four of the L2 participants even completed a bachelor degree in Italy. All HLSs were first exposed to Italian from birth (age 0) except one participant whose age of onset was 6 years, but results did not deviate significantly from the other group members. All HLSs had two native-speaking parents except one who had a native-speaking mother only. In turn, the L2 group’s starting age of first exposure was 13 years. Both the HLSs and L2Ss self-assessed their language skills on a scale from 0 (= no proficiency) to 5 (= native-level proficiency) with respect to their ability in both languages. The average scores for Italian in the HL group were 3.75 in reading, 3.42 in speaking, 3.00 in writing, and 3.75 in listening, while for Swedish these were 5.00 in reading, 5.00 in speaking, 4.92 in writing, and 4.92 in listening. In contrast average scores for the L2 group were 3.08 in reading, 2.75 in speaking, 2.75 in writing, and 3.25 in listening for Italian, in contrast to 4.83 in reading, 4.92 in speaking, 4.83 in writing, and 4.92 in listening for Swedish. The self-assessed proficiency scores in all four skills were highly correlated between the HL and L2 groups (Spearman’s rank-order correlations: reading-speaking, r = .92, p < .001; reading-writing, r = .78, p < .001; reading-listening, r = .99, p < .001; speaking-writing, r = .82, p < .001; speaking-listening, r = .92, p < .001; writing-listening, r = .78, p < .001) and internally consistent (Cronbach α = .94, n = 12), which attests to the strong similarity in high proficiency between the two groups. Thus, the key difference between the HL and L2 group amounted to age of onset and estimated total exposure to Italian which was conducive to appropriately testing the predictions of the FFFH (Table 2).

110 

 Francesco Romano

Table 2: Participant information. Groups

N

HL L2 L1

14 13 18

Age

Proficiency

AFE IT

EE in years

M

Range

M

SD

Range

M

Range

M

Range

32 44 24

16–52 28–54 21–35

71 75 86

12 11 6

57–89 61–95 75–95

0 18 0

– 13–38 –

22 11 24

8–52 2–27 21–35

Note: AFE IT= age of first exposure to Italian. EE= Estimated exposure to Italian in years.

7.2 Materials 7.2.1 Oral structural priming task In order to test the predictions of the MSIH, two tasks imposing different pressure and tapping into different knowledge types were chosen: an oral structural priming task and a timed GJT. Structural priming is a paradigm widely used to define the robustness of a syntactic representation in a speaker’s competence (Bernolet, Hartsuiker & Pickering, 2013; Hartsuiker & Bernolet, 2015; Jackson & Ruf, 2016; Kutta, Kaschak, Porcellini & Jones, 2017; Mahowald, Futrell & Gibson, 2016). The version used in the present study was a variant of Branigan et al. (2005) and subsequent works. Participants first see a picture containing a prime sentence that they were instructed to read out loud (8s timeout) (Figure 2). In this way, a structure was primed both aurally and visually. There followed a fixation point on the screen for 500 ms. and a new blank screen containing a true or false comprehension question related to the picture (3 sec timeout). Subsequently, a new slide containing four prompts for the target sentence appeared above a (new) matching picture. Participants were instructed to use the prompts to form a complete sentence describing the picture, which had to be spoken aloud before the trial timed out (10 s) and the next trial started. In order to test gender, primes and targets were CLLD structures that varied according to verb host as in (8a), (8b), (8c), exemplified also in Figure 2 from left to right. CLLD structures were selected at the expense of simpler structures which may have resulted in speakers at this level of proficiency plateauing. The type of verb is ignored here as it represented the focus of another study (Author, 2020). The targets, represented graphically by the three pictures with prompts in Figure 2, were equally split for objects bearing MASC and FEM gender. A contrast in the object’s feature set of the prime (sigaretta, giornale, and piano at the top of Figure 2) and target sentences (valigia, pesci, and buca at the bottom of Figure 2) was always maintained so that the sets differed by one feature. For instance, if the object of the prime bore MASC and SG features (giornale

Figure 2: Sample trial in the priming task. Photos are taken with permission from the International Picture Naming Project (Szekely et al., 2004), Heaton (1966), and Van Patten, Lee, and Ballman (1992).

Chapter 4 Ultimate attainment of gender in heritage and L2 Italian   111

112 

 Francesco Romano

and piano), the object of the target bore MASC.PL or FEM.SG (pesci and buca). This choice was made in order to avoid facilitating retrieval of the referent’s features in the target by effect of priming. The dislocated objects to which an agreeing clitic had to be chosen belonged to any of the three main noun declension classes shown in Figure 1. The nouns were: italiano, libri, cappotto, vestiti, giardino, tramezzini, muro, Pietro, pacco, piano, biscotto, compito, strumento, fuoco, tavolo, uomo (class I); bollette, sigaretta, cena, palla, mamma, lavagna, casa, lettera, coppa, pasta, pagella, musica, birra, mappa, scarpa, buca, valigia (class II); lezione, giornale, elefante, classe, televisore, pane, campione, fiori, pesci (class III). (8) a. I pesci, Pietro li cucina the fish.OBJ.MASC Pietro them.cl[SG.MASC] cooks.Vfin all’ aperto in-the-outdoors “the fish, Pietro is cooking them outdoors”   (Lexical) b. I pesci, Pietro li vuole cucin-are the fish.OBJ.MASC Pietro them.cl[SG.MASC] want.MOD cook.V all’ aperto in-the-outdoors “The fish, Pietro wants to cook them outdoors”   (Functional) c. I pesci, Pietro li fa cucin-are the fish.OBJ.MASC Pietro them.cl[SG.MASC] make.CAUS cook.V dalla nonna by-the grandma “The fish, Pietro has them cooked by grandma” (Quasi-functional) The primes had structure OBJ SUBJ CL VP PP, differing minimally for the VP phrase. Three versions of each item were created, but all participants saw only one version at a time. There was no lexical overlap between the words in the prime and the prompts for the target sentence, which included, in this order, a bare noun object, a bare noun subject, a verb in the infinitive form, and a fully formed PP. Prime and target verbs were semantically and lexically unrelated as this overlap has been shown to promote priming in L2Ss (Bernolet et al., 2013). Three parallel versions of the task were created, each consisting of 54 prime-target sentence pairs, of which 6 were practice trials, 24 critical items, and 24 fillers. The fillers, held constant across the three versions of the task, primed 24 sentences equally divided between two structures: transitives (k = 12) and passives (k = 12). Prime–target pairs were always separated by a filler trial and conditions alternated. The order of presentation of trials was automatically randomized for each participant. Separate pictures were designed for all prime and target sentences.

Chapter 4 Ultimate attainment of gender in heritage and L2 Italian 

 113

All pictures were black-and-white or grayscale drawings depicting the subject, object, and action of the sentence.

7.2.2 Timed grammaticality judgment task The purpose of selecting a written GJT was to elicit knowledge of gender marking under more explicit conditions in which participants had more time to respond and greater awareness of the structure being tested. The task presented participants items in both grammatical and ungrammatical conditions to be judged as correct or incorrect. One difference to previous versions of this task is that an additional response/button, labelled ‘not sure’, was available. Another difference was that decision times for (un)grammaticality were recorded and responses timed out after 5 seconds. The task consisted of the same 48 target and filler sentences as priming task version 1, with the exception that each target sentence was designed in a grammatical and ungrammatical condition, totaling 96 target items (48 x 2). The targets alternated between two parallel versions of the task so that no participant saw the same item in both conditions in any one version. Each ungrammatical sentence consisted of only one error related to gender in either a SG or PL context (la for lo, and vice versa, li for le, and vice versa) but never an error for number (e.g. lo for li, and vice versa, la for le, and vice versa). Targets additionally varied according to the three verb hosts as in the priming task. The order of presentation of items was automatically randomised for each participant so that fillers and targets alternated and the same verb host type would not appear consecutively. The task began with instructions before three sample items were explained. Eight practice items followed before the experiment began. Participants were informed of the 5-second timeout per response and instructed to press 1 for ‘correct’, 2 for ‘incorrect’, or 3 for ‘not sure’. Sentences appeared all at once. After a response was recorded or a sentence timed-out, there followed a fixation point on the screen for 500 ms and a new blank screen containing a new sentence.

7.3 Procedure Experiments were run on two Lenovo ThinkPad 4173DC9 laptops in designated lab spaces at the University of Venice Ca’ Foscari and Stockholm University. Both tasks were designed and the experiment run on E-Prime experimental software 2.0. Participants’ oral responses in the structural priming task were recorded via the laptops’ in-built microphone and transcribed by the author, a native speaker of Italian. Participants completed the linguistic background questionnaire and place-

114 

 Francesco Romano

ment test online prior to arriving the testing lab for administration of the priming task and GJT, always in that order. The priming task took 33 minutes whilst the GJT lasted on average 5 minutes depending on individual participant speed.

8 Results 8.1 Priming task The FFFH predicts that exposure to the L2 before the critical period and the presence of uGen in the L1 facilitate ultimate attainment of the uGen in the L2. Thus, accuracy levels in the L2 but not the HL group are expected to be significantly lower than the monolingual speakers. To test this prediction, gender accuracy was measured in the priming task. For this analysis, sentences with structure other than the CLLD presented in (8) were excluded as they availed no opportunity for gender agreement. On the other hand, sentences with incorrect co-referents where a clitic referred to the subject rather than the expected dislocated object were retained for counts of gender accuracy (e.g. expected target La mamma, Pietro la saluta prima di uscire, response Pietro la mamma lo saluta prima di uscire where gender agreement is highlighted in boldface). Thirdly, instances of clitic omission were considered for accuracy counts only when they occurred in CLLD structures. Finally, errors in number agreement were disregarded for the purpose of tallying incorrect gender agreement. Table 3 reports the accuracy scores for gender in the priming task. Table 3: Accuracy in use of gender in the priming task. Response

Gender

HL Count

L2 %

Count

L1 %

Count

%

Correct

MASC FEM

91 86

36 34

83 89

33 36

189 166

48 43

Incorrect

MASC FEM

38 37

15 15

41 38

16 15

15 18

4 5

Sum

MASC FEM

129 123

51 49

124 127

49 51

204 184

52 48

252

100

251

100

388

100

Total

Note. HL = heritage group; L2 = L2 group; L1 = monolingual group. Missing data: L1, 44/432 (10 %); HL, 36/288 (12.5%); L2, 37/288 (12.5%).

Chapter 4 Ultimate attainment of gender in heritage and L2 Italian 

 115

As shown in the table, contrary to the FFFH, the HL and L2 groups performed quite similarly, with % accuracy scores of 36 MASC/34 FEM and 33 MASC/36 FEM respectively, well below those of the L1s (48 MASC/43 FEM). The group differences were supported statistically by mixed-effects logistic analyses carried out in R version 3.6 (R Development Core Team, 2013) via the lme4 package. The dependent variable was coded as a binary response representing log odds of correct/incorrect gender agreement where incorrect responses were coded as the sum of omissions plus gender agreement errors. Group was entered as a fixed-effect factor with three levels while participants and items were the random effects. Random slopes were not included in this and subsequent analyses due to sample size, bearing in mind that maximizing the random effects structure does not always entail best fit (Matushek, Kliegl, Vasishth, Baayen & Bates, 2017). An ANOVA type 3 analysis found a main effect of group (X2 = 10.5, df = 2, p < .01). To check which differences between groups were significant, a simple effects analysis with treatment coding was computed by switching the reference level (relevel() function in R) to create the necessary contrasts for each level of the group factor. Pairwise comparisons (Table 4) reveal that the L1s were significantly more likely to produce correct gender agreement than the HLSs and the L2Ss, but the latter did not differ significantly from each other. Table 4: Pairwise comparisons for accuracy in the priming task. Contrasts Intercept L1 vs. HL L1 vs. L2 HL vs. L2

Estimate –1.1713 2.482 2.290 0.193

SE

Wald z

p

.64 .86 .85 .88

–1.818 –2.879 2.669 0.217

.06 .05; L2: X2 = .108, df = 1, p > .05; HL: X2 = .001, df = 1, p > .05). Thus, the patterns so far suggest no clear default form in production against the MSIH and differ from the results previously attested for Spanish.

8.2 GJT To further test the prediction of the FFFH that exposure prior to the critical period and presence of gender in the L1 facilitates ultimate attainment of gender in the L2, we turn to the accuracy scores obtained in the timed GJT (Table 6). Only responses faster than 1000 ms. were considered. The results suggest a strong similarity between the HL and L2 group who score lower than the L1 group. In order to test for the significance of these differences, an ANOVA type 3 analysis was run on the same model fit as the one employed for the priming task. The analysis finds a main effect of group (χ2 (2) = 8.01, p = .01). To check which differences between groups were significant, a simple effects analysis with treatment coding was computed by switching the reference level (relevel() function in R) to create the necessary contrasts for each level of the group factor.

Chapter 4 Ultimate attainment of gender in heritage and L2 Italian 

 117

Table 6: Accuracy in use of gender in the GJT. Response

Gender

HL Count

L2 %

Count

L1 %

Count

%

Correct

MASC FEM

217 186

44 38

227 188

45 38

406 315

52 41

Incorrect

MASC FEM

51 39

10 8

49 35

10 7

34 22

4 3

Sum

MASC FEM

268 225

54 46

276 233

55 45

440 337

56 44

493

100

499

100

777

100

Total

Note. L1, monolinguals. HL, heritage speakers. L2, L2 learners. Missing responses were 14, 16, and 26% from the L1, L2, and HL data.

Pairwise comparisons (Table 7) reveal that the L1s were significantly more likely to produce correct gender agreement than the HLSs and the L2Ss, but the latter did not differ from each other. Contrary to the FFFH, then, L2 and HL groups performed quite similarly and significantly worse than the L1s. Table 7: Pairwise comparisons for accuracy in the GJT. Contrasts

Estimate

SE

Wald z

p

Intercept L1 vs. HL L1 vs. L2 HL vs. L2

–2.2531 1.4352 1.189 0.245

.44 .55 .55 .58

–5.089 2.600 2.144 0.418

.05; L2: χ2 (1) = .092, p > .05; HL: χ2 (1) = .496, p > .05), inconsistent with the MSIH. With respect to response times, the boxplot in Figure 4 suggests a greater difference in mean responses between the MASC and FEM condition (indicated by the centered middle line in each bar) in the HL group than all other groups. To ascertain this conclusion, a mixed-effect linear regression was fit to the data with response time as a continuous dependent. As expected, an ANOVA type 3 finds a significant effect of gender only in the HL group (L1: χ2 (1) = 3.5288, p > .05; L2: χ2 (1) = 1.1576, p > .05; HL: χ2 (1) = 6.9115, p < .01). The main effect was due to the probability of a significantly faster response in the FEM than the MASC condition (β= .3108, SE= .11, t= 2.629, p Essere > Omission. Specifically, the auxiliary avere is pre-

146 

 Maria Teresa Bonfatti-Sabbioni

ferred with unergative and transitive verbs as well as with weather verbs. Essere is used with unaccusative verbs expressing change of location such as andare (to go), partire (to leave), tornare (to come back), entrare (to enter), essere (to be) and verbs expressing a change of state like morire (to die) and nascere (to be born), which specifically encode a telic end point (Sorace, 2000: 867). Only one subject (Subj. Avve) selected essere with transitive verbs guardare (to watch) and scrivere (to write) and with the unergative lavorare (to work). Results show that heritage speakers of Italian syntactically differentiate between unaccusative and unergatives verbs and that they are sensitive to the gradience of unaccusativity identified by Sorace (2000) in her Auxiliary Selection Hierarchy. In fact, most of the subjects, as Italian heritage speakers, exhibit strong intuitions with respect to core unaccusative and unergative verbs of the continuum. Overall, subjects selected essere with unaccusative verbs and avere with unergatives. The following figure shows the degree of unaccusativity decreasing from one endpoint to the other with the target verbs used in the study. Change of location: entrare (to enter), andare (to go), partire (to leave), tornare (to come back); Change of state: nascere (to be born), morire (to die); Existance of state: essere (to be); Weather: nevicare (to snow), grandinare (to hail), piovere (to rain); Uncontrolled process: ridere (to laugh); Controlled motional process: nuotare (to swim), correre (to run Controlled non-motional process: telefonare (to phone), lavorare (to work); Transitive

Figure 1: Unaccusativity gradience with target verbs.

According to the Auxiliary Selection Hierarchy, verbs between the two extremes of the continuum identify an area of variability in auxiliary selection, depending on their distance from the core verbs. This type of verb is located in the middle of the continuum, farther away from both endpoints, where verbs display the highest degree of unaccusativity. The center position provides them with some degree of variation in their auxiliary selection. Both parents (three out of six) and children (three out of seven) allowed for more variation in their judgment of intermediate unaccusative and unergative verbs, located far from the continuum endpoints, displaying an inconsistent pattern in their auxiliary selection for weather verbs. Unergative verbs like lavorare (to work) and correre (to run) express agentive processes in which the subject is not undergoing the action expressed by the verb but represents the agent of the event. They are located at the other end point of the continuum, and they display low degree of unaccusativity and consequently, a high degree of ergativity. Different languages display variation in the auxiliary selection with these verbs, according to the agentive role of the subject. For example, Standard Italian allows

Chapter 5 Auxiliary selection in heritage speakers of Italian  

 147

both auxiliaries with the verb correre (to run). Results from interlanguage analysis highlight the presence of variation in auxiliary selection with weather verbs as well as with unergative correre (to run) and this is identified among members of the same family nucleus and across subjects (Tables 5 and 6). This variation found in the input providers as native speakers of the language of origin is also highlighted by Sorace as being a feature of Standard Italian. Standard Italian allows both auxiliaries with weather verbs when “the telic endpoint of the event is overtly expressed as a change of location” (Sorace 2000: 878), while other languages such as Dutch, French, and German display less variation in the auxiliary selection of weather verbs towards avere (Sorace 2000: 877). Weather verbs fall into the middle of Sorace’s hierarchy allowing the selection of both essere and avere in different languages. Different languages display variability in the auxiliary selection with these verbs, according to the agentive role of the subject. For example, Standard Italian allows both auxiliaries with the verb correre (to run). Auxiliary omission seems to be chosen only 3% of the time with the unergative verb lavorare (to work), with transitive verbs ascoltare (to listen), chiudere (to close), portare (to bring) and with the unaccusative morire (to die). One subject chose the option displaying the auxiliary omission for the verb lavorare. This inherent property of the verb allows for variability in the auxiliary selection within individual languages like French or Dutch. A few subjects also used avere with verbs expressing existence (ES) such as essere (to be), with no change component at all. A similar behavior is identified in monolingual Italian children, who seem to acquire auxiliary selection early in age and free of deviations from the adults’ language (Caprin et al. 2005). Monolingual Italian children display distinction between unaccusative and unergative verbs (Lorusso 2007). Although Lorusso reports accuracy rates in auxiliary selection with unaccusative verbs and lower rates of selection with unergative verbs, deviations are still identified in the child’s language, in the forms of auxiliary omissions or overextensions mostly with variable behavior verbs, which can select both auxiliaries in different syntactic contexts (Devlin, Folli, and Sevdali 2015). The heritage speakers’ behavior in the present study highlights their sensitivity towards the semantics properties of verbs and more specifically, towards the unaccusativity gradient, as identified by Sorace in her Auxiliary Selection Hierarchy model (Sorace 2000) for native monolingual speakers of Italian. In fact, the interdependency between verb semantics and the gradient of unaccusativity is responsible for the verbs ‘auxiliary requirement (Sorace 2000, 2004). The fact that some of the subjects displayed a preference for the auxiliary avere with some unaccusative verbs expressing change of state that usually require essere, is not evidence of absence of native-like attainment (Montrul 2005), but it can be attributed to cross-linguistic influences or to the role of frequency in the use of the heritage language. The results from this study provide answers to the initial research questions as well as evidence in support of the study’s hypothesis:

148 

 Maria Teresa Bonfatti-Sabbioni

RQ1 Are heritage speakers of Italian sensitive to the semantic properties of verbs according to the Auxiliary Selection Hierarchy? (Sorace 2000) A1 Yes. Heritage speakers of Italian make auxiliary selection according to the semantic properties of verbs with respect to the auxiliary selection in the past tense passato prossimo. In addition, results from the present study confirmed the initial hypotheses: H1 Heritage speakers are guided by native intuition in auxiliary selection. With respect to RQ1, results showed that heritage speakers of Italian seem to rely on the semantic properties of the verbs for auxiliary selection according to the Auxiliary Selection Hierarchy (Sorace 2000). In fact, they seem to be guided by a native linguistic sensitivity, which operates within the structures of their heritage grammar. The children as input receivers predominantly behaved in auxiliary selection like the input providers, who are also native speakers of Standard Italian, for most of the target verbs, but not for all. In fact, within Sorace’s continuum of unaccusativity gradient (Sorace 2000), the differences are identified with verbs displaying inherent gradient of variability among languages. Their behavior fit not only the auxiliary selection strategy employed by various native speakers but is also in line with the variation displayed by other Romance languages with the same verbs, whether their selection conforms with rules of the source language or not. The difference in auxiliary selection between heritage speakers and input providers may be viewed as evidence of gradient variation in the auxiliary selection of unaccusative verbs. In addition, it could be hypothesized that heritage grammar differs from the source language in the cutoff point of unaccusative verbs within the continuum of unaccusavity gradient, since the extent of auxiliary selection variation with intransitive verbs, within languages and across languages and varieties “is a function of the position of a verb in the hierarchy” (Sorace 2000: 861). Variation in auxiliary selection is also identified in various Italo-Romance varieties, some of which display a person-driven auxiliary selection (D’Alessandro and Roberts 2010). Whether the differences in auxiliary selection displayed by heritage speakers of Italian is viewed as evidence of gradient variation or as an effect of cross-linguistic influences, I contend that the heritage speakers’ auxiliary selection is a native manifestation of their inbuilt grammar since they display the same variation identified in the source language and identifiable in other languages (Sorace 2000; Donohue 2015).

Chapter 5 Auxiliary selection in heritage speakers of Italian  

 149

11 Extra-linguistic factors Although not specifically investigated in the present study, it is important to mention the role of extra-linguistic factors in the acquisition of a heritage language, such as age of onset acquisition (AOA) of heritage language, length of life in the host country, daily use of heritage language within the family nucleus, exposure to a heritage language community, presence or lack of heritage language instruction and exposure to heritage language literacy. Additional factors defining the social and linguistic profile of the input source, such as the social status of the parents and their level of education, the differences in the parents’ and children’s linguistic behavior, the parents’ intention, and motivation in using the heritage language as well as the children’s response, could also impact heritage language development. Furthermore, interest and language aptitude could also influence the individual development of the family language, favoring in some cases heritage language comprehension over production. In the specific case of this study, the subjects seem to share the same socio-linguistic profile as displayed in the following table (Table 8). Table 8 summarizes the answers gathered from the initial questionnaire administered to the subjects of the study. Most of the questions required a yes/no answer, represented by +/- in the table. Table 8: Linguistic profile of the children as heritage speakers of Italian. Extra-Linguistic Factors Age of HL exposure Length of life in the US Education in HL Education in English Daily use of HL in family Exposure to HL community Daily use of other HL sources

Rom

Isa

Lollo

Cesco

Avve

Elli

G.

0 22+ − + + − −

0 16 − + + − −

0 18 − + + − −

0 15 − + + − −

0 21 − + + − −

0 14 − + + − −

0 13 − + + − −

The subjects represent a homogenous group of heritage language speakers who spent their entire life in the US (the parents’ host country), were exposed to two languages since birth, and manage to use their HL daily in interaction with the input providers. The subjects also share the lack of instruction in the heritage language as well the lack of a community of practice that could favor the use of the heritage language in interaction with speakers other than the family members. These factors could play an important role in the way heritage speakers approach specific domains of grammar where the source language itself displays degrees of variation, such as the semantic properties of verbs in auxiliary selection in Italian language.

150 

 Maria Teresa Bonfatti-Sabbioni

12 Weaknesses of the study This study presents two main methodological challenges: a) The type of task administered to the subjects for data collection; b) The lack of specific criteria of heritage language acquisition. Many experimental linguistic tasks tend to favor the subjects’ metalinguistic awareness and not the implicit knowledge of grammar (Ellis 2005), influencing the speakers’ own perceptions of linguistic correctness. In fact, heritage speakers display an indirect or implicit access to the system of their heritage language since they have been aurally exposed to their family language since birth, often without HL explicit instruction. Therefore, the type and modality of the task administered in this study may have played a role in the way the subjects approached the selection of responses. The stimuli used in the study, although proposed in both oral and written modalities, may lack authenticity in terms of language use. In addition, subjects may have focused more on understanding the text and on providing appropriate answers, rather than naturally performing according to their linguistic intuition. The other methodological challenge resides in the difficulty of identifying criteria of heritage language acquisition: “How do we know what learners know?” (Lakshmanan and Selinker 2001) remains a big challenge in establishing if, and when a given structure has been acquired. One way to overcome this challenge is to identify alternative diagnostics for determining the heritage speakers’ competence (Lakshmanan and Selinker, 2001). For example, we could explore the acquisition of heritage grammar through an order of acquisition of given elements, since various levels of competence in the family language could be identified according to the order in which specific elements are acquired by the speakers.

13 Conclusion This study provides evidence for the claim that the heritage language as family language can be investigated as an independent linguistic system with its own set of rules displaying structural differences from the language of origin. In fact, the differences in auxiliary selection between the children as heritage input receivers and the parents as heritage input providers identify systematic differences that occur at the level of grammar between the heritage language and the source language. These differences may not be triggered by reduced properties of input, but by the specific nature of the input provided and therefore by the specific nature of the source language. In fact, variation in the heritage system seems to take place in those domains of grammar particular to the source language, and in which the

Chapter 5 Auxiliary selection in heritage speakers of Italian  

 151

source language displays degrees of variation itself. Therefore, variation displayed by heritage speakers doesn’t identify the presence of vulnerable grammatical domains in the heritage system, but instead reflects the presence of domains of variation in the structures of the source language. To conclude, the analysis of the structural properties of heritage grammar and how they are acquired, may lead us to question what we assume and know about first language acquisition in monolingual settings. Therefore, I believe that by identifying the grammatical domains of contrast and variation in the source language, it is possible to predict the areas of heritage grammar in which re-analysis and re-structuring take place. Further studies in various fields of linguistics are needed to uncover mechanisms underlying the heritage system. Typological investigations could contribute to determining if the heritage system shares similarities with other world languages, according to universal rules or to cross-linguistic influences, or if it is the product of individual conditions of language acquisition, meaning a heritage grammar for each heritage speaker. Additional investigations on the syntactic and semantic properties of the heritage system may shed light on new aspects in grammatical theory.

References Belletti, Adriana & Maria Teresa Guasti. 2015. The acquisition of Italian: morphosyntax and its interfaces in different modes of acquisition. Philadelphia: John Benjamins Publishing Company. Belletti, Adriana, Elisa Bennati & Antonella Sorace. 2007. Theoretical and developmental issues in the syntax of subjects: Evidence from near-native Italian. Natural Language & Linguistic Theory 25(4). 657–689. Bley-Vroman, Robert. 1983. The comparative fallacy in interlanguage studies: the case of systematicity. Language Learning 33(1). 1–17. Cook, Vivian (ed.). 2003. Effects of the second language on the first (vol.3). Bristol: Multilingual Matters. Cuza, Alejandro & Frank Joshua. 2011. Transfer effects at the syntax semantics interface: The case of double-que questions in heritage Spanish. The Heritage Language Journal 8(2). 66–89. Cuza, Alejandro. 2012. Crosslinguistic influence at the syntax proper: Interrogative subject– verb inversion in heritage Spanish. International Journal of Bilingualism 17(1). 71–9. Devlin, Megan, Raffaella Folli, Allison Henry & Christina Sevdali. 2015. Clitic right dislocation in English: Cross-linguistic influence in multilingual acquisition. Lingua 161. 101–124. D’Alessandro, Roberta & Ian Roberts. 2010. Past participle agreement in Abruzzese: Split auxiliary selection and the null-subject parameter. Natural Language & Linguistic Theory 28(1). 41–72. Devlin, Meghan, Raffaella Folli & Christina Sevdali. 2015. The lexicon-syntax interface in 3L1 context: an experimental investigation of unaccusative/unergative distinction in English and Italian. Crosslinguistic Influence and Crosslinguistic Interaction in Multilingual Language Learning, New York: Bloomsbury Publishing Inc.

152 

 Maria Teresa Bonfatti-Sabbioni

Donohue, Cathryn. 2015. On variable auxiliary selection in Western Europe. Linguistic Research 32(2). 469–480. Ellis, Nick C. 2005. At the interface: Dynamic interactions of explicit and implicit language knowledge. Studies in second language acquisition 27(2). 305–352. Ellis, Nick C. 2008. The dynamics of second language emergence: Cycles of language use, language change, and language acquisition. The modern language journal 92(2). 232–249. Kupisch, Tanja. 2006. The emergence of article forms and functions in the language acquisition of a German-Italian bilingual child. In Conxita Lleó (ed.), Interfaces in multilingualism. 139–177. Filiaci, Francesca. 2011. Anaphoric preferences of null and overt subjects in Italian and Spanish: A cross-linguistic comparison. Edinburgh: The University of Edinburgh Dissertation. La Morgia, Francesca. 2016. Assessing the relationship between input and strength of language development: A study on Italian-English bilingual children. In Jeanette Treffers-Daller & Carmen Silva-Corvalan, Language Dominance in Bilinguals. Issues of Measurement and Operationalization, 195–218. Cambridge: Cambridge University Press. Lakshmanan, Usha & Larry Selinker. 2001. Analyzing interlanguage: How do we know what learners know? Second Language Research 17 (4). 393–420. Lepschy, Anna Laura & Guilio Lepschy. 2013. The Italian Language Today (2nd edn). Oxford: Routledge Publishing Ltd. Lorusso, Paolo, Claudia Caprin & Maria Teresa Guasti. 2005. Overt subject distribution in early Italian children. A supplement to the Proceedings of the 29th annual Boston University Conference on Language Development (BUCLD 29), Boston University Conference. Lorusso, Paolo. 2007. The acquisition of aspect in L1 Italian. Proceedings of the 2nd Conference of Generative Approaches to Language Acquisition North America (GALANA), 253–264. Somerville: Cascadilla Proceedings Project. Montrul, Silvina. 2008. Incomplete acquisition in bilingualism: Re-examining the age factor (Vol.39). Philadelphia: John Benjamins Publishing. Montrul, Silvina. 2010. Current issues in heritage language acquisition. Annual Review of Applied Linguistics 30. 3–23. Montrul, Silvina. 2012. Bilingualism and the heritage language speaker. In Tej K. Bhatia & William C. Ritchie (eds.), The handbook of bilingualism and multilingualism. 168–189. New Jersey: Wiley and Sons. Rothman, Jason. 2007. Heritage speaker competence differences, language change, and input type: Inflected infinitives in Heritage Brazilian Portuguese. International Journal of Bilingualism 11(4). 359–389. Rothman, Jason. 2009. Understanding the nature and outcomes of early bilingualism: Romance languages as heritage languages. International journal of bilingualism 13(2). 155–163. Polinsky, Maria. 2006. Incomplete acquisition: American Russian. Journal of Slavic linguistics 14(2). 191–262. Romano, Francesco. 2020. Ultimate attainment in heritage language speakers: Syntactic and morphological knowledge of Italian accusative clitics. Applied Psycholinguistics 41(2). 347–380. Rosi, Fabiana & Jessica Cancila. 2007. The acquisition of Italian Tense-Aspect morphology: Toward a native-like strategy. Studi e Saggi Linguistici. XLV. 169–189. Scontras, Gregory, Zusana Fuchs & Maria Polinsky. 2015. Heritage language and linguistic theory. Frontiers in Psychology (6). 15–45. Selinker, Larry. 1972. Interlanguage. IRAL – International Review of Applied Linguistics in Language Teaching 10(1–4). 209–232.

Chapter 5 Auxiliary selection in heritage speakers of Italian  

 153

Sorace, Antonella. 2000. Gradients in auxiliary selection with intransitive verbs. Language Vol.76 (4). 859–890. Sorace, Antonella. 2004. Native language attrition and developmental instability at the syntaxdiscourse interface: Data, interpretations, and methods. Bilingualism: Language and Cognition 7.2. 143–145. Sorace, Antonella and Francesca Filiaci. 2006. “Anaphora resolution in near-native speakers of Italian.” Second language research 22.3. 339–368. Sorace, Antonella and Ludovica Serratrice. 2009. Internal and external interfaces in bilingual language development: Beyond structural overlap. International Journal of Bilingualism 13.2. 195–210. Sorace, Antonella, Ludovica Serratrice, Francesca Filiaci & Michela Baldo. 2009. Discourse conditions on subject pronoun realization: Testing the linguistic intuitions of older bilingual children. Lingua 119.3. 460–477. Tsimpli, Ianthi and Antonella Sorace. 2006. Differentiating interfaces: L2 performance in syntax, semantics, and syntax-discourse phenomena. In David Bamman, Tatiana Magnitskaia & Colleen Zaller (eds.), Proceedings of the Annual Boston University Conference on Language Development, Morocco, 2005, 653–664. Somersville: Cascadilla Press. Valdés, Guadalupe. 2014. Expanding definitions of giftedness: The case of young interpreters from immigrant communities. Oxford: Routledge Publishing Ltd. Wiley, Terrence G., Kreeft Peyton, Christian Joy, Donna Moore, Sarah Catherine K, Liu Na, (eds.). 2014. Handbook of Heritage, Community, and Native American Languages in the United States: Research, Policy, and Educational Practice. Oxford: Routledge Publishing Ltd.

Jacopo Torregrossa, Irene Caloi, Andrea Listanti

Chapter 6  The acquisition of syntactic structures in heritage Italian: Assessing the role of language exposure at critical periods Abstract: The present study aims to show that language exposure at certain age ranges is crucial for the successful acquisition of certain syntactic structures, such as complement clauses, adverbial clauses, clauses involving movement and clauses involving both movement and embedding. Twenty-nine Italian heritage children living in Germany and ranging in age between 7;08 and 12;06 took part in the study. We assessed their acquisition of syntactic structures in Italian by using a Sentence Repetition Task. We also assessed their language exposure across different age ranges (between 0 and 3, 3 and 6 and at 6 years) by using parental questionnaires. The results show that the acquisition of different syntactic structures is sensitive to different language exposure variables as related to different age ranges. We interpret the results as showing that the effects of language exposure on the acquisition of syntactic structures are modulated by their timing in monolingual language acquisition. Language exposure in critical periods seems to play a relevant role in bilingual language acquisition. Keywords: acquisition of syntax; embedding; movement; sentence repetition task; language exposure; critical period

1 Introduction The present study aims to investigate to what extent the acquisition of different types of syntactic structures by Italian heritage children in Germany is affected by language exposure variables. We would like to offer a nuanced view of the relationship between the acquisition of linguistic structures and the amount of exposure that the children receive in Italian. In particular, we investigate to what extent the acquisition of certain syntactic structures is sensitive to the amount of exposure children received at the time in which the target structure is supposed to emerge in monolingual language acquisition. In other words, we consider the effects of language exposure variables on the acquisition of a structure in relation to the timing of monolingual acquisition of the structure itself. https://doi.org/10.1515/9783110759587-007

156 

 Jacopo Torregrossa, Irene Caloi, Andrea Listanti

Across the literature, several studies on bilingual language acquisition have shown that the amount of exposure to a certain language affects the timing and nature of its acquisition. For example, Unsworth (2013) shows that cumulative language exposure to Dutch predicts accuracy in the production of grammatical gender in Dutch by English-Dutch bilingual children. Cumulative language exposure is intended as exposure to a certain language across different contexts (family, school, extra-curricular activities) over time (at present and in the past). The impact of cumulative language exposure on language acquisition has been confirmed by other studies dealing with the acquisition of other linguistic phenomena by bilinguals of other language combinations (see, e.g., Dijkstra et al. 2015 on vocabulary acquisition in Dutch and Frisian by Dutch-Frisian bilingual children and Bedore et al. 2012 on semantic and morphosyntactic acquisition in English by English-Spanish bilingual pre-kindergarteners). However, some studies dealing with the acquisition of other linguistic phenomena have not found the same effect of cumulative language exposure on heritage language acquisition (see Romano 2020 for a similar consideration). For example, Repetto (2010) shows that German-French bilingual children ranging in age between 1;6 and 3;6 produce very few word-order errors in both languages, thus behaving like their respective monolingual peers, in spite of a lower amount of language exposure (see also Schmeißer et al. 2015). In the domain of vocabulary, the study by Thordardottir (2011) reveals that English-French bilinguals exhibit the same performance in receptive vocabulary as their monolingual peers, even if they receive 40–60% of language exposure compared to the latter. Tsimpli (2014) notices that the linguistic phenomena that are not sensitive to cumulative language exposure are the ones that tend to emerge early in monolingual language acquisition. The evidence reported in Schulz and Grimm (2019) is consistent with this hypothesis: the authors show that simultaneous bilingual children with German as one of their languages are as accurate as their German monolingual peers in the production and comprehension of early acquired phenomena like subject-verb agreement and the expression of telicity. By contrast, linguistic phenomena that emerge late or very late in monolingual acquisition tend to be affected by cumulative language exposure (see Tsimpli 2014 for discussion). These include bilingual children’s acquisition of gender marking in Dutch (as discussed above; Unsworth 2013), negation and case marking in German (Schulz and Grimm 2019) and marked word orders in Italian (Listanti and Torregrossa submitted). In a recent study, Torregrossa et al. (2021) show that cumulative language exposure is a strong predictor of the type of referring expressions that Greek heritage children tend to use in narrative production. Notably, the conditions of use of referring expressions in discourse is acquired very late in monolingual language acquisition.

Chapter 6 The acquisition of syntactic structures in heritage Italian 

 157

Tsimpli (2014) accounts for the observed differences in timing of monolingual acquisition in terms of the distinction between “core” syntactic structures and interface phenomena. While the former (core structures) involve only narrow syntax and are “semantically vacuous both in terms of triggers and effects” (Tsimpli 2014: 285), the latter require the integration of information across several domains, such as syntax and discourse or syntax and language-external cognitive resources. In this contribution, we will not distance ourselves from Tsimpli’s idea, although we agree with Schulz and Grimm (2019) that many other factors may affect timing of monolingual acquisition, including, for example, “how much idiosyncrasy and irregularity is involved in a construction” (Schulz and Grimm 2019: 4). In addition to timing of monolingual acquisition, another factor that may interact with cumulative language exposure in bilingual children’s acquisition of a certain phenomenon is crosslinguistic influence from their other language. Cross-linguistic influence may be associated with acceleration or delay of the acquisition process (Müller 2017). Let us consider a situation in which a bilingual’s two languages pattern similarly with respect to a certain phenomenon. If  this phenomenon emerges early in one of the bilingual’s languages, this might result in an accelerated acquisition of the same phenomenon in the other language as compared to monolingual children (see, e.g., Genesee 2001 on the acquisition of finiteness in English by French-English bilinguals; Kupisch 2007 on the acquisition of gender-marking in German by Italian-German bilinguals; Liceras, Fuertes, and de la Fuente 2012; Torregrossa and Bongartz 2018). By contrast, if a bilingual’s two languages differ from each other with respect to a certain linguistic phenomenon, interference of one language on the other may occur, which may also lead to a delay in the acquisition process (see, e.g., Müller and Hulk 2001). By looking at bilingual children’s acquisition of different types of syntactic structures, we may be able to contribute to the debate on how the interaction between timing in monolingual language acquisition and measures of cumulative language exposure affects heritage language acquisition. As we will show in the next section, syntactic structures differ from each other in their degree of complexity, which is reflected in their different timing of monolingual acquisition. Furthermore, the syntactic structures that we will consider in this paper exhibit a different degree of “cross-linguistic robustness”, at least for the language combination considered here, i.e., German and Italian. The two languages pattern similarly to each other in certain properties exhibited by syntactic structures, while they exhibit a different behavior in other properties.

158 

 Jacopo Torregrossa, Irene Caloi, Andrea Listanti

2 Bilingual acquisition of syntactic structures: The role of timing of monolingual acquisition, language exposure and cross-linguistic effects The acquisition of syntactic structures requires a certain degree of cognitive maturity, which varies depending on how complex the syntactic computation of each structure is. In this contribution, we consider complement clauses, adverbial clauses, structures generated by movement and structures generated by movement and embedding (see Tsehaye et al. 2021 for a similar selection of structures). In the next subsections, we will consider each structure in turn, from both a syntactic and a cross-linguistic perspective.

2.1 Complement clauses A complement clause is an embedded clause which is selected by a verb, functioning as its argument. In a recently published corpus study on L1 Italian, Moscati and Rizzi (2021) show that both finite (i.e., introduced by che ‘that’) and non-finite complement clauses (i.e., introduced by the complementizers di and a  – ‘to’ in English) emerge by the age of 3. The early emergence of complement clauses has been observed in other languages beyond Italian (see, e.g., Vasilyeva, Waterfall, and Huttenlocher 2008 on English and Brandt, Lieven, and Tomasello 2010 on German). This observation is consistent with Tsimpli’s (2014) hypothesis that “core” structures are acquired relatively early, given that embedding via complementation is a fundamental operation in narrow syntax (Section 1). In a recent study on Italian heritage children in Greece ranging in age between 8 and 12, Andreou, Torregrossa, and Bongartz (2021) have shown that the accurate production of complement clauses is affected by the amount of language exposure that the children received between 0 and 3 years and at the age of 6. This observation suggests that mastery of a certain structure is sensitive to language exposure variables related to the critical period in which the structure is supposed to emerge in monolingual acquisition (i.e., between 0 and 3 in this case). The observed impact of exposure variables related to age 6 on children’s accuracy in the production of complementation comes as a surprise. However, the authors claim that this effect is triggered by the acquisition of other aspects of complementation, such as the use of the right complementizer in non-finite clauses. This may be particularly challenging for Greek-Italian bilingual children, given that Greek does not allow for non-finite clauses and there is no systematic pattern motivating the use of di vs. a in Italian (see the above reference to idiosyncrasy in Schulz and Grimm 2019). Cross-linguis-

Chapter 6 The acquisition of syntactic structures in heritage Italian 

 159

tic effects may be observed in the present study too. On the one hand, Italian complement clauses differ from German ones in that they are not verb-final.1 However, bilingual children should be accurate in their production of word order, given that children acquire the word-order possibilities which are available in a language relatively early. On the other hand, the accurate selection of complementizers in non-finite clauses should be challenging for Italian heritage children with German as their other language too, especially given that German allows for only one complementizer in non-finite clauses (i.e., zu ‘to’).

2.2 Adverbial clauses Adverbial clauses are “adjoined” to the main clause and modify its meaning, just like adverbs. Several studies have characterized adverbial clauses as syntax-discourse interface structures. They are typically used to establish discourse coherence and cohesion, connecting sentences to each other (Thompson, Longacre, and Hwang 2007). For example, while-clauses have the function of introducing background information for a preceding (or following) sentence. Thus, the acquisition of adverbial clauses requires the integration of lexical and morphosyntactic knowledge (e.g., the meaning of connectives, word order in subordinate clauses, marking of temporal and aspectual relations between clauses) as well as knowledge of the discourse conditions for their appropriate use. Although subordinating connectives seem to emerge relatively early in monolingual language acquisition (i.e., around the age of 3; see, e.g., Diessel 2004), children may exhibit difficulties in the interpretation of some types of adverbial clauses even by the age of 8 (see, e.g., the study by Pyykkönen, Niemi, and Järvikivi 2003 on the comprehension of before- and after-clauses; van Veen 2011 for discussion). This finding is in line with Tsimpli’s hypothesis that syntax-discourse interface structures tend to emerge late or (very) late in language acquisition. Along the same lines, Andreou, Torregrossa, and Bongartz (2021) have analyzed the acquisition of adverbial clauses by Greek-Italian bilingual children ranging in age between 8 and 12 in both their languages. They showed that knowledge of adverbial clauses is sensitive to bilingual children’s current exposure to their respective languages, providing evidence that the age range between 8 and 12 is crucial for the acquisition of this kind of clauses.

1 German also allows for complement clauses exhibiting verb-second, but only in the absence of a complementizer (cf. Brandt, Lieven, and Tomasello 2010).

160 

 Jacopo Torregrossa, Irene Caloi, Andrea Listanti

With respect to the language combination at issue in this paper, it should be mentioned that German adverbial clauses exhibit verb final, whereas this is not the case of Italian ones. However, German-Italian bilingual children should be able to acquire this language-specific property relatively early, as has been observed with complement clauses. Connectives should also emerge early in both languages, as discussed above. By contrast, mastery of certain aspectual forms in Italian adverbial clauses may pose some challenges to German-Italian bilingual children. For example, Italian while-clauses are associated with the use of imperfective forms if they refer to events in the past. Unlike Italian, German does not mark aspectual distinctions with past verb forms. Therefore, German-Italian bilingual children may take some time to figure out how the aspectual system in the two languages works (see Diaubalick and Guijarro-Fuentes 2016 for similar considerations on Spanish-German adult bilinguals).

2.3 Structures generated by movement In this study, we consider also structures generated via movement. Crucially, these structures do not involve embedding (contrary to the structures considered above), but may be nevertheless associated with some degree of syntactic complexity (Rothman et al. 2016). In particular, we consider object clitic left dislocations (CLLDs), passive sentences and object wh-questions. Object CLLDs are generated via movement of the object constituent to the left periphery of the sentence. The constituent is resumed by a clitic pronoun in sentence-internal position (Rizzi 1997; Torregrossa 2012). Belletti and Manetti (2019) show that by the age of 4, Italian children master the syntax of CLLDs (i.e., movement and clitic resumption) as well as the discourse conditions for their appropriate use (see also Moscati and Rizzi 2021). However, they may encounter difficulties with intervention configurations in which the object constituent moves across an overtly expressed subject, even at later stages of acquisition (see Section 4.2 for some relevant examples). Along the same lines, production and comprehension experiments reveal that Italian children master passive sentences (including by-phrases) around the age of 4 (Manetti 2013; Volpato, Verin, and Cardinaletti 2016). However, the overt realization of by-phrases as well as the use of passive sentences with non-actional verbs may be problematic until children reach the age of 6 (see Belletti and Guasti 2015 for discussion). As for the acquisition of object wh-questions in Italian, Guasti, Branchini, and Arosio (2012) show that both their comprehension and production may be problematic for 5-year-old children, whose accuracy does not exceed 75%. As in the case of object CLLDs, children’s difficulties are mainly related to the intervention configuration involving movement of the object constituent

Chapter 6 The acquisition of syntactic structures in heritage Italian 

 161

across an intervening subject (see Belletti and Guasti 2015). Overall, all reviewed studies suggest that movement configurations are late-acquired structures, since they require children to develop sensitivity to the discourse conditions affecting their use as well as a certain degree of cognitive maturity necessary to cope with complex intervention configurations. Bilingual children may need more time to acquire these structures as compared to their monolingual peers (Schulz and Grimm 2019). On the one hand, structures requiring the integration of syntactic and discourse information have been shown to be particularly challenging in bilingual language acquisition (see, e.g., Sorace 2011 as a main reference). On the other hand, cross-linguistic effects may modulate the acquisition process and outcomes related to these structures. For example, topicalizations involve movement and clitic resumption in Italian, but only movement in German (Frey 2004). Figuring out which form-function possibilities are available in each language may delay the acquisition process.

2.4 Structures generated via embedding and movement Our analysis will also consider structures generated via both embedding and movement, including subject and object relative clauses as well as finite and non-finite indirect interrogatives. The above considerations related to the acquisition of movement structures apply to this last group of structures too. Furthermore, the derivation of these structures involves embedding too, which is likely to be associated with increased syntactic complexity. For example, the production and comprehension of relative clauses in which the object constituent moves across an overtly realized subject has been shown to be challenging even for Italian children between 5 and 6 years (e.g., Belletti et al. 2012). By contrast, subject relative clauses tend to emerge at the same time (or slightly later) as complement clauses (i.e., between 2 and 3 years). The nature of the moved constituent (e.g., argument vs. adjunct) is very likely to modulate the difficulty in the production and comprehension of structures involving movement and embedding. Once again, cross-linguistic differences as well as the higher or lower sensitivity of different structures to discourse constraints may contribute to variation in the timing of acquisition of these structures among bilingual children (see, e.g., Brandt et al. 2009 on the effects of discourse in the comprehension of relative clauses). Structure involving movement and embedding differ significantly between the two languages considered in this paper. First, they all require verb-final in German (unlike in Italian), because they involve embedding (see above). In the case of relative clauses, case-marking is a cue for argument structure (e.g., distinguishing between the subject and the object of an object relative clause) in German, but

162 

 Jacopo Torregrossa, Irene Caloi, Andrea Listanti

not in Italian. In Italian object relative clauses, the identification of the subject is usually based on structural cues: the subject of an object relative clause usually corresponds to the full noun that follows the complementizer and precedes the embedded verb.2

3 The study The present study investigates Italian heritage children who were exposed to German (the majority language in the society) between their birth and 6 years of age and were attending German-Italian bilingual schools in Berlin or Hamburg (Germany) at the moment of testing. Our study involved two tasks: – background questionnaires for parents, in order to identify patterns of language use at home, with particular attention to children’s exposure to Italian (Section 4.1), – a Sentence Repetition Task in Italian (SRT, henceforth) as a measure of children’s language proficiency (Section 4.2). We intend to investigate to what extent Italian heritage children’s acquisition of different syntactic structures is sensitive to the amount of language exposure at a specific age range, focusing on language exposure between 0 and 3, 3 and 6 and at 6 years. This will be taken as evidence that a specific age range is crucial for the acquisition of a certain phenomenon by heritage children, along the lines of what has been shown for monolingual language acquisition (see Section 2). 2 However, the subject of an object relative clause can also occur in postverbal position. For example, both (i) and (ii) can be interpreted as object relative clauses (as is made clear by the English translation of both sentences). However, (ii) is ambiguous between an object relative clause featuring a postverbal subject and a subject relative clause in which the mother is the object. In the German object relative clause in (iii), the young boy corresponds to the object of the relative clause (and the mother to the subject) because the relative pronoun (which refers to the young boy) bears accusative case. (i)

(ii)

(iii)

Il bambino che la mamma the boy that the mother ‘The boy that the mother is greeting’ Il bambino che saluta the boy that greets ‘The boy that the mother is greeting’ Der Junge, den die Mutter the boy.NOM.SG. REL_PRO.ACC.SG. the mother ‘The boy that the mother is greeting’

saluta greets la mamma the mother grüßt greets

Chapter 6 The acquisition of syntactic structures in heritage Italian 

 163

3.1 The predictions of the study Based on the studies related to the timing of monolingual acquisition of different syntactic structures (Section 2), we make the following predictions. First, the acquisition of complement clauses should be sensitive to exposure measures related to children’s early years, i.e., between 0 and 3 or slightly later, depending on whether the simultaneous acquisition of German will slow down the acquisition of Italian or not. Second, the acquisition of adverbial clauses should be sensitive to exposure measures related to children’s later years, i.e., after 6. Finally, as for the acquisition of structures involving movement and movement combined with embedding, two scenarios are possible. On the one hand, we may observe that heritage children’s acquisition of these structure is sensitive to language exposure in a specific age range, which is either the same as the one observed among monolingual children or a slightly later one (Section 2). On the other hand, we may observe that the acquisition of these structures is sensitive to different age ranges. For example, knowledge of structures involving embedding and movement requires knowledge of embedding (which may emerge relatively early) and movement (which may be acquired later). Likewise, knowledge of CLLDs implies knowledge of clitics coupled with the command of movement operations. Clitics seem to emerge relatively early in both monolingual and bilingual language acquisition (e.g., around the age of 3 in Schaeffer 2000 or even before, according to Schmitz and Müller 2008). In other words, knowledge of structures involving movement and both movement and embedding is “constructed” across the years, which may be reflected on the sensitivity of these structures to different language exposure measures, corresponding to different critical age ranges.

3.2 Participants We tested 29 German-Italian heritage children ranging in age from 7;08 to 12;06 (M: 10;03; SD: 1;07), 11 living in Berlin (age range: 9;02–12;06; M: 11;04; SD: 1;00) and 18 in Hamburg (age range: 7;08–10;09; M: 8;10; SD: 1;00). All children were exposed in school to both Italian and German at the time of testing. The children in Berlin attended a German-Italian bilingual primary school (from year 1 to 6). The children in Hamburg attended a primary school with a bilingualer Zweig (i.e., ‘bilingual branch’ or ‘sector’), where the teaching of Italian involved both language and content (e.g., history and geography) from the first year onwards (Caloi and Torregrossa 2021). Before the study, the parents provided written informed consent. The children were simultaneous bilinguals (N: 13), being exposed to both German and Italian

164 

 Jacopo Torregrossa, Irene Caloi, Andrea Listanti

from birth, early successive ones (N: 2), whose first exposure to German was at age 3, or early second language learners (N: 8), being first exposed to German when they entered primary school at the age of 6. Six questionnaires did not contain the relevant information related to children’s first exposure to German. Both the parents and the teachers reported that none of the children had previously identified speech or hearing impairment.

4 Materials and methodology 4.1 Background questionnaires The parents were administered an extensive questionnaire targeting information related to children’s use of one or the other language from birth until the age of 6 and across different contexts (home and school; Bongartz and Torregrossa 2020; Torregrossa et al. 2021). The questionnaire was administered in written form in Italian and German, so that the parents were able to answer in the language in which they felt more comfortable.3 Using questionnaire information is common practice in bilingualism research (see Kašćelan et al. 2022 for a complete overview), although the reliability of parental reports has been criticized in some studies (e.g., Marchman et al. 2017). For the purposes of the present study, we focused on the questions related to children’s language history. This is based on the hypothesis that the acquisition of different syntactic structures is sensitive to different exposure variables as related to the time in which each syntactic structure is acquired in monolingual acquisition (Section 2). Parents were asked to provide information on the languages used in and outside of the home. The questions addressed three different stages: before the child entered kindergarten (i.e., between 0 and 3 years), when the child attended kindergarten (between 3 and 6 years) and when the child entered primary school (at the age of 6). For example, for the age range between 0 and 3 years, we asked “Do you remember which languages your child heard and used from birth until the age of three?”. The parents were provided a list of persons who may have interacted with the child (mother, father, siblings, grandparents and baby-sitters; for the age range 3–6 and the age of 6, we also added reference to teachers and classmates). They had to indicate whether each person spoke “mainly Italian”, “mainly German” or “both

3 The questionnaire included no question concerning whether the parents were speakers of an Italian dialect or whether this dialect was used at home.

Chapter 6 The acquisition of syntactic structures in heritage Italian 

 165

languages” to the child. Here below we report the table in the questionnaire related to the question concerning language exposure between 3 and 6. Table 1: Table from the parental questionnaire related to the question “which languages did your child hear and speak when s/he attended kindergarten (between 3 and 6)?”. mainly German

both languages

mainly Italian

Mother Father

Sibling 1

Sibling 2 Sibling 3

Grandparents Baby sitter

Kindergarten teachers Kindergarten friends

For each person speaking “mainly Italian” or “mainly German” to the child, one point was assigned to the corresponding language. For answers reporting that the person used “both languages”, we split the point between the two languages. This allowed us to derive two separate scores, one for German and one for Italian. Then, we calculated the ratio between the language specific score and the total score corresponding to the age range at issue (i.e., the total amount of persons interacting with the child). Thus, for each stage (0–3, 3–6 and 6), the child was associated with two scores, one for Italian and one for German, ranging from 0 (no exposure to the language) to 1 (exposure only to the language). For the rest of our analysis, we will consider the score indicating children’s exposure to Italian. Before proceeding to the next section, two considerations are in order. First, the questions reported in the questionnaire do not consider the amount of time that each person spent with the child. This choice was mainly related to the observation reported in previous studies that the richness and diversity of the input received by a child affects acquisition outcomes more than quantitative measures of language exposure (Blom and Soderstrom 2020; Gollan, Starr, and Ferreira 2015 and Torregrossa, Flores, and Rinke 2022). Second, “grandparents”, “kindergarten teachers” and “kindergarten friends” were considered collectively, in order to avoid making the questionnaire too long and giving too much weight to these persons compared to the persons within the closest family circle. However, it was useful to add reference to these persons, in order to gain insight into how bilingual the environment of the family (outside the closest circle) and the kindergarten was.

166 

 Jacopo Torregrossa, Irene Caloi, Andrea Listanti

4.2 Sentence Repetition Task (SRT) We used the Italian SRT designed by Torregrossa, Eisenbeiß, and Bongartz (2022), which was modeled on the Greek SRT in Andreou, Torregrossa, and Bongartz (2021). Participants had to listen to a sentence and repeat it right afterwards. Performance in a SRT usually activates both comprehension and production resources: after hearing a sentence, participants are supposed to analyze its meaning, reconstruct it and encode it for output (Klem et al. 2015; Marinis and Armon-Lotem 2015). Therefore, SRTs have often been indicated as reliable instruments for assessing participants’ syntactic proficiency. Furthermore, it has been shown that bilingual children’s performance in a SRT seems to correlate with other measures, including vocabulary knowledge and language exposure variables (Andreou, Torregrossa, and Bongartz 2021). Differently from more “traditional” SRTs, in which the target sentences are independent from each other, the SRT used in this study involves a discourse dimension: the sentences are connected to each other to create a story. The story is about an alien whose spaceship breaks and falls on the earth and a beaver who tries to help him. By repeating each target sentence, the participant contributes to the development of the story-plot. The narrative dimension is beneficial at different levels: the unfolding of a plot engages children’s attention, the reoccurrence of characters and objects reduces the cost for lexical activation, and the narrative context appears more ecological for the elicitation of structures involving the syntax-discourse interface (e.g., object relative clauses or clitic left dislocation) – see Torregrossa, Eisenbeiß, and Bongartz (2022). Target sentences were prerecorded in a studio by a female Italian native speaker and presented on a laptop. After hearing and repeating each sentence, the child was shown the corresponding picture (Figure S1 in Appendix). The SRT consists of a total of 28 sentences including complement clauses, adverbial clauses, structures involving movement of the object constituent and argument movement combined with embedding (e.g., relative clauses or indirect interrogatives). Table S1 in Appendix reports a list of all sentences included in the SRT. We made sure that all sentences had similar length in terms of number of syllables. The task includes 7 complement clauses, either of the finite type, as introduced by the complementizer che (‘that’; see (8) and (21’) in Table S1 and (1) reported below), or of the non-finite type and introduced by either di or a (see (3’), (8’), (11), (26), (27) in Table S1 and (2) reported below), depending on the functional element selected by the finite verb in the main clause.

Chapter 6 The acquisition of syntactic structures in heritage Italian 

 167

(1) L’alieno gli dice che la navicella The alien CLIT.DAT.M.3SG say.PRS.3SG that the spaceship ha smesso di funzionare AUX.PRS.3SG ceased to work ‘The alien tells him that the spaceship has ceased to work’ (2) Li raccoglie e prova a rimetterli CLIT.OBJ.M.3PL collect.PRS.3SG and try.PRS.3SG to put-CLIT.OBJ.M.3PL insieme con un cacciavite together with a screwdriver ‘He collects them and tries to put them back together with a screwdriver’ We also included 4 adverbial clauses, i.e., subordinate clauses expressing a temporal meaning (see (12) in Table S1 and (3) below), a causal meaning (see (3) and (1) in Table S1 and (4) below) or a final meaning (see (18) in Table S1). (3) Dopoché ha messo i pezzi insieme, accende il motore After AUX.PRS.3SG put the pieces together, start.PRS.3SG the motor ‘After he has put the pieces together, he turns the engine on’ (4)

L’alieno è tanto triste perché non è The alien BE.PRS.3SG very sad because NEG AUX.PRS.3SG riuscito a guidarla managed to drive-CLIT.OBJ.F.3SG ‘The alien is very sad because he has not managed to control it’

Furthermore, the SRT includes 8 structures involving sub-extraction and movement of the object constituent from its canonical argument position. This is the case of passives, in which the object is promoted to subject position for verb agreement (see (13), (19) and (24) in Table S1 and (5) below). The movement can alternatively result in object fronting to the clausal left periphery, as in the case of object wh-questions (see (15) and (28) in Table S1) or clitic left dislocations, which also involve an obligatory resumptive pronoun in the clause (see (9), (14) and (20) in Table S1 and (6) below). (5) Il povero animale viene circondato improvvisamente da un The poor animal surround.PASS.PRS.3SG suddenly by a fumo nero smoke black ‘The poor animal is suddenly surrounded by black smoke’

168  (6)

 Jacopo Torregrossa, Irene Caloi, Andrea Listanti

Alcuni pezzi, l’alieno li ha persi durante Some pieces, the alien CLIT.OBJ.M.3PL AUX.PRS.3SG lost during la caduta the crash ‘Some of the pieces, the alien has lost them during the crash’

The last group of target structures combines argument movement and embedding and consists of 10 sentences. This is the case of subject and object relative clauses, in which the embedded clause is derived via movement of one of its arguments (the one corresponding to the head of the relative clause) out of its canonical position (see (1), (6), (17), (18’) and (25) in Table S1 and (7) below for subject relatives, and (5), (10), (23) and (8) below for object relatives). Furthermore, we included two indirect interrogatives, one finite (see (7) in Table S1) and one not-finite (see (4) in Table S1). (7) Il castoro accende la navicella che prima sembrava The beaver start.PRS.3SG the spaceship that before seem.PST.IPFV.3SG distrutta ruined ‘The beaver starts the spaceship, which seemed completely ruined before’ (8)

Il castoro trova i pezzi che l’alieno stava cercando The beaver find.PRS.3SG the pieces that the-alien search.PST.IPFV.PROG.3SG ‘The beaver finds the pieces, that the alien was searching’

4.3 Procedure Children were tested at school. At the beginning of the session, they were told they were about to listen to a story about a beaver and an alien, but they would hear only one sentence at the time. If the children wanted to hear the following sentence and find out how the story moved forward, they had to repeat the sentence they had just heard. After giving the children instructions, there was a practice session, in which children could ask questions and the experimenter provide feedback, e.g., explain the task again or encourage the children to repeat the sentences (cf. Marinis and Armon-Lotem 2015). The task was administered in form of a Power Point Presentation. Children could listen to each sentence only once. After repeating the sentence, they were shown the corresponding picture. Then, the target sentence in the following slide was played. Children always received positive feedback, independently of whether they had managed to repeat the target sentence correctly. If they were not able to repeat a sentence, they were told not to worry and that they

Chapter 6 The acquisition of syntactic structures in heritage Italian 

 169

could move forward to the following picture. Sentence repetitions were recorded and later transcribed by the experimenter for the analysis and scoring.

4.4 Data analysis Children received one point for each target structure that they were able to repeat. As long as the target structure was repeated correctly (without errors), the point was assigned, independently of whether the rest of the sentence was incomplete or ungrammatical, or the child used different lexical items in the repetition.4 The substitution of the target structure with another structure (e.g., the substitution of a passive sentence with an active one) was considered as an incorrect repetition (see Marinis and Armon-Lotem 2015 for this methodology). We decided to analyze the data in this way, based on previous studies showing that the correct production of the target structures of a SRT is a more reliable indicator of participants’ syntactic proficiency than the correct verbatim repetition of sentences or the production of grammatical sentences (see, e.g., Hamann and Abed Ibrahim 2017). For each participant, we calculated four different scores corresponding to the amount of complement clauses, adverbial clauses, clauses involving movement and movement combined with embedding that were correctly produced. Since the maximum score that could be achieved varies based on the type of structure (e.g., 7 with complement clauses and 4 with adverbial clauses), we expressed each score as a percentage of the maximum possible score corresponding to the type of structure at issue. This was done in order to compare children’s performance across structures.

5 Results 5.1 Background questionnaires Figure 1 reports the distribution of the scores related to children’s exposure to Italian from different persons in and outside of the home across different age stages (from birth to 3 years, from 3 to 6 years and at 6 years), as extracted from the 4 For instance, the repetition of (8) in Table S1 was considered as correct if the child said: L’alieno gli dice che la navicella ha smesso (‘The alien said to him that the spaceship has stopped’) or L’alieno gli dice che la navicella ha finito (‘The alien said to him that the spaceship has finished’) – see Torregrossa, Eisenbeiß, and Bongartz (2022) for additional methodological details.

170 

 Jacopo Torregrossa, Irene Caloi, Andrea Listanti

parental questionnaires. Children’s exposure to Italian decreases over time from the mean value of .72 (SD = .32) between 0 and 3 years, to .66 (SD = .30) between 3 and 6 years and .54 (SD  = .26) when children were 6 years old. A decrease of exposure to the heritage language is commonly observed among heritage speakers, especially after the age of 3, when exposure to the societal language increases due to fact that children usually enter the kindergarten. 1.00

Italian_input

0.75

0.50

0.25

0.00 0-3

3-6

6

age_range Figure 1: Boxplot of the distribution of the scores related to children’s exposure to Italian from different persons in and outside of the home across different age stages (0–3, 3–6 and 6 years). The scores are expressed as a percentage of the number of persons from which the child received input in Italian over the total number of persons from which the child received input in one or the other language (see Section 4.1). The values range from 0 (no exposure to Italian) to 1 (exclusive exposure to Italian).

5.2 Analysis of the SRT Before presenting the results related to each type of structure, we report in Table 2 the different target structures sorted out into the abovementioned four different types. The number associated with each sentence corresponds to the one provided in Table S1. The part in bold corresponds to the target structure considered in the analysis.

Chapter 6 The acquisition of syntactic structures in heritage Italian 

 171

Table 2: List of all target structures considered in the analysis as sorted out based on the type of structure (Please see Table S1 for translations). Complement clauses

Adverbial clauses

Clauses involving movement

Clauses involving movement and embedding

(3’) L’alieno è tanto triste perché non è riuscito a guidarla.

(3) L’alieno è tanto triste perché non è riuscito a guidarla.

(9) Alcuni pezzi, l’alieno li ha persi durante la caduta.

(1) C’era una volta un alieno che andava sulla Terra con la sua navicella.

(8) L’alieno gli dice che la navicella ha smesso di funzionare.

(12) Dopoché ha messo i pezzi insieme, accende il motore.

(13) Il povero animale viene circondato improvvisamente da un fumo nero.

(4) Inoltre non sa proprio come rimettere insieme i pezzi e ripararla.

(8’) L’alieno gli dice che la navicella ha smesso di funzionare.

(18) per salvarlo dalla guerra interna che lo sta distruggendo.

(14) Ma il motore, il castoro non riesce ad accenderlo.

(5) Arriva un castoro che un cacciatore sta inseguendo nella foresta.

(11) Li raccoglie e prova a rimetterli insieme con un cacciavite.

(21) perché ha visto che è un animale dal cuore d’oro.

(15) Chi stanno cercando gli alieni sul nostro pianeta e perché?

(6) Il castoro vede l’alieno che disperato cerca aiuto.

(21’) perché ha visto che è un animale dal cuore d’oro.

(19) Gli alieni saranno aiutati da un animale della Terra.

(7) Lo raggiunge e gli chiede cosa mai gli sia successo.

(26) Decide di salirci e seguirlo nello spazio.

(20) Il castoro, l’alieno lo vuole portare con sé sul suo pianeta.

(10) Il castoro trova i pezzi che l’alieno sta cercando.

(27) Alla fine, i due amici iniziano a viaggiare felici.

(24) La navicella può essere guidata di nuovo dall’alieno.

(17) Cerchiamo animali gentili che vengano sul nostro pianeta

(28) Quali alieni conoscerà il castoro nello spazio?

(18’) per salvarlo dalla guerra interna che lo sta distruggendo. (23) Lui guarda ammirato il motore che il castoro ha riparato. (25) Il castoro accende la navicella che prima sembrava distrutta.

172 

 Jacopo Torregrossa, Irene Caloi, Andrea Listanti

5.2.1 Accuracy scores across different structure types Figure 2 reports the means (and ± 1.5 standard errors) for the response accuracy associated with each type of structure (complement clauses, adverbial clauses, clauses involving movement and movement combined with embedding, respectively). A Kruskal-Wallis test (R function: kruskal.test()) shows that response accuracy differs across the four clause types (χ2(3) = 35.78, p < .001).5 A Dunn’s test for pairwise comparisons (R function: dunnTest(); package FSA, see Ogle et al. 2021) reveals that this difference is due to the fact that the clauses involving movement and embedding have a lower response accuracy than the other three types of clauses (vs. adverbial clauses: z = 5.67, p