Developing sound systems in trilingual first language acquisition (Standard German, Swiss German and Italian) – The acquisition of consonant inventories, stops and gemination Doctoral thesis for obtaining the academic degree of Doctor of Philosophy (Dr. phil.) submitted by Marieke Josephine Einfeldt at the Faculty of Humanities Department of Linguistics Konstanz, 2021 Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-2-v14kydptv4b35 Date of the oral examination: 29 April 2022 First referee: Prof. Dr. Tanja Kupisch Second referee: Dr. Mechtild Tronnier Third referee: Prof. Dr. Bettina Braun A b s t r a c t | I ABSTRACT This dissertation investigates the development of the sound system(s) of a balanced trilingual (3L1) child acquiring Standard German, Swiss German and Italian simultaneously. Trilingual first language (3L1) acquisition still is an under-researched area of multilingual first language acquisition and the effects of reduced input and interaction between three first languages (L1s) are not yet widely explored (Chevalier, 2011; Hoffmann, 2001a; Hoffmann, 2001b; Kupisch, 2023). The third language adds an additional layer of complexity to trilingual as compared to bilingual language acquisition. The relative input in each language is more reduced, and there is an extra source for cross- linguistic influence (CLI), i.e., systematic interaction between a multilingual’s first languages. The few existing studies on phonetics and phonology in 3L1 acquisition suggest language separation with CLI (Mayr & Montanari, 2015; Montanari, 2011a; Yang & Hua, 2010). In the present dissertation, three segmental phenomena were examined in three empirical studies: (I) consonant inventories, (II) stop patterns and (III) gemination. The results are discussed in the light of earlier discussions on CLI and language separation in multilingual first language acquisition (Paradis & Genesee, 1996). There are three types of CLI: Acceleration (i.e., earlier emergence of a phenomenon compared to monolingual acquisition), deceleration (i.e., later emergence of a phenomenon compared to monolingual acquisition) and transfer (i.e., transfer of a phenomenon from one language to the other). The role of language internal (e.g., markedness) and external (e.g., input quantity) factors that have been shown to be source of CLI in bilingual language acquisition (Kehoe, 2015) are examined more closely in 3L1 context. An additional CLI factor that will be discussed is typological proximity. This expands the set of factors that can be investigated in the context of 3L1 acquisition, as a direct investigation of typological proximity effects is only possible if there are more than two languages involved. The three studies in this dissertation are based on naturalistic data that was produced by a trilingual child collected in a longitudinal design between the ages of 1;4 and 2;6. Study I examines the development of consonant inventories. The sounds in the merged trilingual consonant inventory may be grouped into three categories based on II | A b s t r a c t their (non)similarities across the three languages: (i) fully shared sounds (e.g., /m/), (ii) partially shared sounds (i.e., sounds not sharing all features such as voicing, e.g., /v/ (Italian and Standard German) vs. /v̥/ (Swiss German)) and (iii) language-specific sounds (e.g., the Italian sound /ʎ/). It is conceivable that these different groups might lead to different outcomes in terms of CLI, e.g., deceleration of language-specific sounds due to reduced input compared to monolingual children or markedness, earlier acquisition of shared sounds due to their higher input frequency, or facilitative transfer of sounds for reasons of economy and articulatory ease. Besides potential CLI effects, particularities of the regional varieties that the child acquires were taken into consideration, e.g., a palatalization process of /s/ to /ʃ/ in the Alemannic variety of Standard German (Auer, 1990). Results suggest a monolingual-like development in the trilingual child with separate language systems and little CLI. There is potential evidence for acceleration in the Standard German data (e.g., /ʃ/), possibly caused by the higher frequency of the sound in the regional variety the child is acquiring. Another explanation is enhanced discrimination ability (Bosch & Sebastián-Gallés, 2001), which has been suggested in previous studies to improve the productions of certain sounds in multilingual children (Montanari, 2011a). Furthermore, transfer was found (e.g., the German and Swiss German sound /h/ in Italian). On a whole, however, the development of the consonant inventories was monolingual-like. Study II examines the development of the trilingual child’s stop patterns. While all languages have binary stop contrasts, they base the distinctions on different parameters. German and Italian differentiate stops based on Voice Onset Time (VOT) but use different values. German distinguishes lenis short-lag sound (e.g., /b/; [b̥]) from fortis long-lag sounds (e.g., /p/; [pʰ]), while Italian lenis sounds are prevoiced (e.g., /b/; [b]), and fortis sounds have a short-lag VOT (e.g., /p/; [p]) (Lisker & Abramson, 1964). All Swiss German sounds, by contrast, have short-lag VOT (e.g., /b̥/; [b̥] and /p/; [p]). The distinctive feature is closure duration (CD): lenis sounds (e.g., /b̥/) have shorter CD than fortis sounds (e.g., /p/) (Ladd & Schmid, 2018). The fact that all three binary stop distinction patterns differ in their realization is ideal for investigating language separation. Furthermore, the factor ‘structural ambiguity’ (i.e., an overlap on the surface structure between languages that leads to structural ambiguity) can be investigated, as the languages share the short-lag feature on the surface, but the feature belongs to different phonological categories, making the interpretation ambiguous. Results of the stop productions’ analysis in all three languages show monolingual-like development and A b s t r a c t | III language separation. The trilingual child’s productions in all three languages have different VOT patterns. The Swiss German CD distinction was neither acquired by the monolinguals nor by the trilingual child. This means the development in all three languages was target-like and none of the CLI predictors had an influence. The acquisition of three different languages did not seem to have an effect on the development of stops, not even in the case of the closely related languages Swiss German and Standard German. Metalinguistic awareness and the aforementioned discrimination abilities seem supportive of language-specific realizations of the stop patterns as well. Study III examines the development of the consonant length contrast in Italian. Consonant length is not distinctive in German, while it is a distinctive feature in Swiss German and Italian. However, consonant length is bound to different phonotactic constraints in the two languages. In Italian word-medial and intervocalic contexts, consonant duration distinguishes short consonants (singletons) from long consonants (geminates; Loporcaro, 1996). The majority of Italian consonants can form semi-minimal pairs based on consonant length such as /pala/ ’shovel’ vs. /palːa/ ‘ball’. In Swiss German, consonant length is a feature of the fortis-lenis-contrast of fricatives and stops (e.g., /p vs. b̥/, Fleischer & Schmid, 2006) and length is distinctive in almost all contexts. A secondary cue to gemination in Italian is vowel duration. Vowels preceding singletons are longer than vowels preceding geminates. Vowel length is contrastive in both Germanic languages. In German, vowel length additionally correlates with vowel quality (only /a/ and /ɛ/ in some varieties are unarguably exclusively duration-bound), while in Swiss German all full vowel qualities can form semi-minimal pairs. This structural overlap might be expected to cause CLI in consonant and vowel duration in Italian geminates: the production of long consonants could be problematic due to CLI from German, or mistakenly be correlated with the fortis-lenis-contrast due to CLI from Swiss German; vowel duration could be more pronounced due to Germanic influence. Results show a monolingual-like or even accelerated development of the consonant length contrast in Italian. Vowel duration development was monolingual-like as well. The development of /a/ and /ɛ/ could hint at CLI from German, as /a/ and /ɛ/ behave differently than all other vowels. However, an unambiguous effect of German is not verifiable as the length contrast is target-like. Furthermore, the fortis-lenis-contrast in Swiss German intervocalic word-medial stops is acquired in a target-like manner. This means that the phonetic duration values were produced correctly within the phonologic length categories. It may also be the case that the potential effects of structural overlap and reduced input were IV | A b s t r a c t arguably counteracted by the child’s higher degree of metalinguistic awareness and discrimination abilities. The present dissertation brings new insights to the understudied area of phonetic and phonological 3L1 acquisition. It presents evidence for the monolingual-like acquisition of the segmental phenomena investigated, namely the consonant inventories, stop patterns and gemination. Results suggest separate sound systems with little CLI. Phonetic features are not only acquired but they are also mapped correctly onto the phonological categories of the respective language. Reasons for this monolingual-like development could be advanced production skills resulting from advanced discrimination abilities and higher metalinguistic awareness. Furthermore, the child is a balanced trilingual, i.e., he has no dominant language. This might be another major factor that explains the child’s monolingual-like development. The study covered an early stage of acquisition. Potentially, the young age correlates with balance as the quantity of input and output are more evenly-spread than later in acquisition. None of the other factors (input, structural ambiguity, frequency, typological proximity and markedness/complexity) appeared to affect the trilingual child’s language acquisition, nor did they lead to differences between the trilingual speaker and the monolinguals. A c k n o w l e d g e m e n t s | V ACKNOWLEDGEMENTS Back in 2009 when I began studying Italian, I never imagined there’d come a time where I’d find myself completing a PhD thesis in linguistics, but here I am, writing the acknowledgements. I am infinitely grateful for the great company throughout this journey and would like to take the opportunity to thank the many wonderful people who have made this experience all the more special. First and foremost, I would like to thank my first supervisor Tanja Kupisch for without her, this thesis would never have been written. I met Tanja in my first-ever linguistic lecture in Hamburg and since then she has accompanied and encouraged me every step of the way. Thank you, Tanja, for your help and support, comments and feedback, believing in me and for introducing me to the academic world. I also want to thank my second supervisor, Mechtild Tronnier. I met Mechtild during my first conference in Arhus and I was more than happy that she agreed to co-supervise my PhD thesis. Thank you, Mechtild, for your advice and critical questioning during the writing and analyses process and your motivating words during our supervision sessions. Thanks also to Bettina Braun, my third supervisor. Bettina taught my first seminar here in Konstanz. As I could not register before the semester started, I had asked her in the first session if I could still join the course. She jokingly said: “As long as you smile and nod like you did this week you can stay.” Thank you, Bettina, for accepting to co-supervise this thesis (even though I did not smile and nod the whole time), for your comments on methodology, your help with statistics and Praat scripting and for always being supportive and optimistic. A very warm and special thank you to ‘Tommaso’ and his parents. I feel so fortunate to have conducted this study with you. Thank you for everything – this PhD thesis would literally not have been the same without you. I always felt so welcome and had the best time, both during the recordings and while working on them. ‘Tommaso’ you are such a cool kid and you have amazing parents! Grazie, Merci und Dankeschön von ganzem Herzen. I want to express my gratitude also to the monolingual children taking part in the study. I am also thankful to the Department of Psychology Developmental Psychology: Infancy and Childhood “Kleine Weltentdecker” at the University of Zurich, especially to Anja Gampe, Marco Bleiker and Nicoletta Dimou, without whom I would not have been able to have collected the Swiss German data. Thank you also to Marie-Anne Morand for VI | A c k n o w l e d g e m e n t s helping me with the Swiss German data. Furthermore, I want to thank Teresa Guasti who paved the way for the data collection in Italy and put me in contact with the Servizi Educativi (Coordinamento Pedagogico) in Piacenza. Thank you also to Bettina Braun for allowing me to use the database of the BabySpeechLab in Konstanz and to Monika Lindauer for helping with the recruitment and answering all my questions. Thanks also to the research assistants who helped with transcribing and annotating the data: Anne-Sophie Hufer, Chiara Ochsenreiter, Paulina Speck and especially Denise Esen and Luisa Dautner. During my PhD in Konstanz, I learnt a lot but more importantly, I met wonderful people. Thanks to my fellow romanistic PhD students Carmen Widera and Katharina Kaiser for warmly welcoming me in Konstanz and for walking this road together. Thank you, Erlinde Meertens, for making my life a little less business-casual and a little more peanut butter sauce and for always being a true friend. Thank you, Ramona Wallner, for being a great friend and for always having an open office door for me. Thank you, Felix Frühauf, for coffee, lunch breaks and table tennis. Thank you, Anika Lloyd-Smith, for being a great office mate and for your advice and help throughout this time from beginning to end. Thank you, Angeliki Golegos, for being an amazing friend and bringing Hamburg to Konstanz. Thank you, Monika Lindauer, for walks by the lake and through the Hockgraben. I am very grateful for being part of P6. Thank you, Nicole Dehé and Bettina Braun, for this opportunity, for always being supportive and giving me time to work on my dissertation. Thank you, Daniela Wochner and Angela James, for amazing teamwork and support. You are truly fabulous. A very special thank you to Katharina Zahner-Ritter for reading parts of this dissertation, helping with statistics and basically any other (non-) linguistic related problem and question. A special thanks to the growing multilingualism community in Konstanz and abroad and everyone else in the department for always having an open ear, being up for a coffee, Romanist*innen auf dem Weihnachtsmarkt and your thoughts on abstracts and presentations. Thank you especially to Andrea Beltrama, Alexandra Rehn, Anna Czypionka, Elisabeth Süß, Fernanda Barrientos, Georg Kaiser, Grazia di Pisa, Kaja Gregorc, Katerina Kalouli, Lari-Valteri Suhonen, Maria Ferin, Miriam Geiss, Sergio Soares, Simon Hellerich, Stefano Quaglia, Sonja Gumbsheimer, Theo Marinis and Svenja Schmid. A c k n o w l e d g e m e n t s | VII I want to express my cordial thanks to Carmen Kelling for always having an open ear and handling administrative problems so well. I want to thank Melanie Hochstätter for being an amazing colleague during my time as coordinator for internationalization, also Tania Simeoni and Irene Wolke for help with organizational issues. Thank you, Max Canzi, for help with the data analysis and (so many) stats questions and Achim Kleinmann for technical support. A very big thank you to Angela James, Anika Lloyd-Smith and Jonathan Hugh for reading the dissertation, comments and corrections – you rock! I want to express my gratitude also to my friends from back home and around the world. Thank you, Miriam, Jule, Moritz, Mim, Jo, Sabrina and Lara – long distance is not always easy but we make it work! Thanks also to my non-linguist friends and roommates here in Konstanz, Katrin, Hanna, Martina, Alina and Orla. You have made my life here so much more fun. Finally, I want to thank my family and especially my parents, Rita and Thomas, for supporting my decisions, wanting me to be happy and always ‘having my back’. I would have never finished this dissertation without you. VIII | A c k n o w l e d g e m e n t s C o n t e n t s | IX CONTENTS Abstract .............................................................................................................................. I Acknowledgements .......................................................................................................... V Contents .......................................................................................................................... IX List of Abbreviations .................................................................................................... XV List of Figures ............................................................................................................. XVII List of Tables ............................................................................................................... XXI 1. Introduction ................................................................................................................... 1 2. Language separation and cross-linguistic influence in early multilinguals .................. 7 2.1. Differences between bilingual and trilingual language acquisition scenarios ....... 7 2.2. Multilingual language acquisition and language separation .................................. 8 2.3. Predictors of Cross-linguistic Interaction (CLI) .................................................. 10 2.3.1. External factors .............................................................................................. 10 2.3.1.1. Input ............................................................................................................ 10 2.3.1.2. Language dominance .................................................................................. 13 2.3.2. Internal factors ............................................................................................... 16 2.3.2.1. Complexity or markedness of a phenomenon ............................................ 16 2.3.2.2. Structural ambiguity ................................................................................... 17 2.3.2.3. Frequency (of occurrence) .......................................................................... 18 2.3.3. A “new” factor: Typological proximity ........................................................ 19 2.4. Terminological note: Dialects or languages? ....................................................... 22 3. Data Collection ........................................................................................................... 25 3.1. Trilingual Child .................................................................................................... 25 3.1.1. Language profile ............................................................................................ 25 3.1.2. Recording setting ........................................................................................... 26 3.1.3. Language use during recordings sessions and dominance ............................ 27 3.2. Control groups: Monolingual children ................................................................. 29 X | C o n t e n t s 3.2.1. Monolingual German children ....................................................................... 30 3.2.1.1 Recruitment, age and regional origin ........................................................... 30 3.2.1.2. Material and data collection ........................................................................ 31 3.2.2. Monolingual Italian children .......................................................................... 31 3.2.2.1. Recruitment, age and regional origin .......................................................... 31 3.2.2.2. Material and data collection ........................................................................ 32 3.2.3. Monolingual Swiss German children ............................................................. 33 3.2.3.1. Recruitment, age and regional origin .......................................................... 33 3.2.3.2. Material and data collection ........................................................................ 33 4. Consonant inventories ................................................................................................. 35 4.1. The German consonant inventory ......................................................................... 35 4.1.1. Regional variation in the south of Baden-Wuerttemberg .............................. 39 4.2. The Italian consonant inventory ........................................................................... 39 4.2.1. Regional variation in Sicily ........................................................................... 42 4.3. The Zurich German consonant inventory ............................................................. 43 4.4. Comparison between the consonant phoneme inventories ................................... 45 4.5. Development and acquisition of the consonant inventory .................................... 49 4.5.1 Monolingual acquisition ..................................................................................... 49 4.5.1.1. German ........................................................................................................ 51 4.5.2. Italian ............................................................................................................. 53 4.5.3. Swiss German ................................................................................................ 56 4.5.4. Comparison of the monolingual Standard German and Italian development 57 4.6. Multilingual first language acquisition ................................................................. 60 4.7. Case study: Consonant Inventories ....................................................................... 68 4.7.1. Research hypotheses ...................................................................................... 69 4.7.2. Method and analyzed material ....................................................................... 70 4.7.3. Results of the acquisition of the consonant inventories ................................. 74 C o n t e n t s | XI 4.7.3.1. Control group: Monolingual German data ................................................. 75 4.7.3.2. Control group: Monolingual Italian data .................................................... 77 4.7.3.3. Control group: Monolingual Swiss German data ....................................... 79 4.7.3.4. Trilingual child: Standard German, Italian and Swiss German data .......... 83 4.7.3.4.1. Trilingual child: German data ................................................................. 86 4.7.3.4.2. Trilingual child: Italian data ................................................................... 87 4.7.3.4.3. Trilingual child: Swiss German data ....................................................... 87 4.7.4. Discussion of monolingual and trilingual data .............................................. 89 4.7.4.1. Control group: Monolingual data ............................................................... 89 4.7.4.2. Trilingual child: Discussion of the trilingual results .................................. 92 4.7.4.2.1. Trilingual German data ............................................................................ 93 4.7.4.2.2. Trilingual Italian data .............................................................................. 94 4.7.4.2.3. Trilingual Swiss German data ................................................................. 95 4.7.5. Discussion: Cross-linguistic influence and general discussion ..................... 96 4.7.6. A Note on methodology .............................................................................. 101 4.7.7. Conclusion ................................................................................................... 102 5. Stop patterns ............................................................................................................. 103 5.1. Defining the concept of Voice Onset Time (VOT) ............................................ 104 5.2. Stop Categories in German, Swiss German and Italian ..................................... 106 5.2.1. Stop categories in German ........................................................................... 106 5.2.2. Stop categories in Italian ............................................................................. 107 5.2.3. Stop categories in Zurich German ............................................................... 109 5.3.4. Comparison of the three binary systems ..................................................... 110 5.3. Acquisition of VOT ........................................................................................... 112 5.3.1. Monolingual VOT acquisition ..................................................................... 112 5.3.2. Multilingual VOT acquisition ..................................................................... 115 5.4. Acquisition of CD (and long consonants) .......................................................... 120 XII | C o n t e n t s 5.5. Case Study: Stop patterns ................................................................................... 122 5.5.1. Research hypotheses .................................................................................... 124 5.5.2. Method ......................................................................................................... 124 5.5.3. Analyzed material ........................................................................................ 127 5.5.3.1. Control group: Swiss German monolingual dataset (VOT) ..................... 127 5.5.3.2. Trilingual child: dataset (VOT) ................................................................ 127 5.5.3.3. Control group: Swiss German monolingual dataset (CD) ........................ 129 5.5.3.4. Trilingual dataset (CD) ............................................................................. 130 5.5.4. Results .......................................................................................................... 130 5.5.4.1. VOT .......................................................................................................... 131 5.5.4.1.1. Control group: Swiss German monolinguals ......................................... 131 5.5.4.1.2. Trilingual child ....................................................................................... 133 5.5.4.2. CD ............................................................................................................. 137 5.5.4.2.1. Control group: Swiss German monolinguals ......................................... 137 5.5.4.2.2. Trilingual child ....................................................................................... 139 5.5.5. Discussion .................................................................................................... 142 5.5.5.1. Control group: Swiss German monolinguals ............................................ 143 5.5.5.2. Trilingual child .......................................................................................... 144 5.5.6. Conclusion ................................................................................................... 149 6. Consonant Gemination .............................................................................................. 151 6.1. Durational features of Italian .............................................................................. 152 6.1.1. Speech rate ................................................................................................... 153 6.1.2. Manner of articulation .................................................................................. 155 6.1.3. Regional variation ........................................................................................ 157 6.2. Durational features of German ........................................................................... 159 6.3. Durational features of Swiss German ................................................................. 159 6.4. Difference in Italian, German and Swiss German consonantal length ............... 161 C o n t e n t s | XIII 6.5. Monolingual language acquisition of geminates ................................................ 162 6.6. Geminates in multilingual language acquisition ................................................ 165 6.7. Case study: Gemination ..................................................................................... 170 6.7.1. Research hypotheses .................................................................................... 171 6.7.2. Method ......................................................................................................... 172 6.7.3. Analyzed material ........................................................................................ 174 6.7.3.1. Control group: Italian monolingual dataset (consonants) ........................ 174 6.7.3.2. Trilingual child: Italian dataset (consonants) ........................................... 174 6.7.3.3. Trilingual child: Swiss German dataset (consonants) .............................. 176 6.7.3.4. Monolingual control group: Italian monolingual dataset (vowel) ............ 176 6.7.3.5. Trilingual child: Italian dataset (vowel) ................................................... 177 6.7.4. Results ......................................................................................................... 178 6.7.4.1. Length contrast in consonants .................................................................. 179 6.7.4.1.1. Control group: Italian monolinguals ..................................................... 179 6.7.4.1.2. Trilingual data: Italian ........................................................................... 182 6.7.4.1.3. Trilingual data: Swiss German stops ..................................................... 184 6.7.4.2. Duration contrast in vowels ...................................................................... 187 6.7.4.2.1. Control group: Italian monolinguals ..................................................... 187 6.7.4.2.2. Trilingual data ....................................................................................... 189 6.7.3. Discussion ................................................................................................... 192 6.7.3.1. Control group: Italian monolinguals ........................................................ 192 6.7.3.2. Trilingual data .......................................................................................... 193 6.7.4. Conclusion ................................................................................................... 197 7. General discussion .................................................................................................... 199 7.1. Summary of empirical results ............................................................................ 199 7.2. Discussion of the results in the light of CLI predictors ..................................... 201 7.2.1. External factors ............................................................................................ 201 XIV | 7.2.1.1. Input .......................................................................................................... 201 7.2.1.2. Dominance ................................................................................................ 204 7.2.2. Internal factors ............................................................................................. 205 7.2.2.1. Complexity or markedness ....................................................................... 205 7.2.2.2. Structural ambiguity .................................................................................. 207 7.2.2.3. Frequency (of occurrence) ........................................................................ 208 7.2.3. Typological proximity ................................................................................. 209 7.3. General (dis)advantages for the sound system through trilingualism? ........... 210 7.4. Limitations of the study .................................................................................. 211 8. Conclusion ................................................................................................................. 213 9. Outlook ...................................................................................................................... 215 Reference list ................................................................................................................. 217 Zusammenfassung ......................................................................................................... 235 Appendix ............................................................................................................................ i Questionnaires cross-sectional study (monolinguals) .................................................... i Standard German ........................................................................................................ i Swiss German ............................................................................................................. i Italian ........................................................................................................................ ii Additional Figures ....................................................................................................... iii Additional Table ......................................................................................................... vii List of German test items .............................................................................................. ix List of Italian test items ............................................................................................... xv List of Swiss German test items ................................................................................. xix L i s t o f A b b r e v i a t i o n s | XV LIST OF ABBREVIATIONS 2L1 Simultaneous Bilingual(ism) 3L1 Simultaneous Trilingual(ism) CD Closure Duration CLI Cross-Linguistic Interaction ELM Early-, Middle, Late-developing Sounds L1 First Language L2 Second Language MLU Mean Length of Utterance PPC Percent Consonant Correct VOT Voice Onset-Time XVI | L i s t o f A b b r e v i a t i o n s L i s t o f F i g u r e s | XVII LIST OF FIGURES Figure Caption Page Figure 1 Language use, language mixing and choice of language based on the language spoken by the interlocutor (Language of the interlocutor from left to right: German, Italian, Swiss German), split by age of the trilingual child. 27 Figure 2 Development of Mean Length of Utterance (MLU) in words over time. 29 Figure 3 VOT continuum with in Italian as an example for true voice languages and German as an example for aspirating languages and Swiss German (no voicing contrast). 104 Figure 4 VOT in German (Left panel: short-lag; right panel: long-lag). 106 Figure 5 Italian VOT (Left panel: prevoicing; right panel: short-lag VOT). 108 Figure 6 Stops in Zurich German (Left panel: short closure; right panel: long closure). 109 Figure 7 Example measurements for CD and VOT. 125 Figure 8 Monolingual Swiss German control group: VOT productions split by age (cross sectional); dashed line indicates dividing line between short-lag and long-lag; red squares indicate mean; fortis stops are represented in blue and lenis stops in pink. 131 Figure 9 Monolingual Swiss German control group: VOT development grouped in developmental stage “young” and “old”; dashed line indicates dividing line between short-lag and long-lag (Stoehr et al. 2018); red squares indicate mean; fortis stops are represented in blue and lenis stops in pink. 132 Figure 10 Trilingual child: VOT development split by language over time; dashed line indicates the dividing line between short-lag and long- lag; red squares indicate the mean; fortis stops are represented in blue and lenis stops in pink. 134 XVIII | L i s t o f F i g u r e s Figure 11 Trilingual child: VOT values split into developmental stages “young” (1;10-2;4) and old (2;4-2;8; dashed line indicates dividing line between short-lag and long-lag; red squares indicate mean; fortis stops are represented in blue and lenis stops in pink. 136 Figure 12 Monolingual Swiss German control group: CD production normalized by speech rate; red squares indicate mean; fortis stops are represented in blue and lenis stops in pink. 138 Figure 13 Monolingual Swiss German control group: Normalized CD divided by developmental stage; red squares indicate mean; fortis stops are represented in blue and lenis stops in pink. 139 Figure 14 Trilingual child: CD (normalized) development over time split by language and raddoppiamento fonosintattico (no = ITALIAN/yes = ITALIAN (RS)); red squares indicate mean; fortis stops are represented in blue and lenis stops in pink. 140 Figure 15 Trilingual child: CD (normalized) development over time split by language and raddoppiamento fonosintattico (no = ITALIAN/yes = ITALIAN (RS)) and grouped by “young” and “old” using 2;4 as cut off point; red squares indicate mean; fortis stops are represented in blue and lenis stops in pink. 141 Figure 16 Monolingual Italian control group: Development of normalized consonant duration split by age. Red squares indicate mean; black dots indicate outliners. Geminates are represented in green; Singletons are represented in pink. 180 Figure 17 Monolingual Italian control group: Development of normalized consonant duration split by age and manner of articulation. Red squares indicate mean; black dots indicate outliners. Geminates are represented in green; Singletons are represented in pink. 181 Figure 18 Trilingual child: Development of normalized consonant duration split by age. Red squares indicate mean; black dots indicate outliners. Geminates are represented in green; Singletons are represented in pink. 182 L i s t o f F i g u r e s | XIX Figure 19 Trilingual child: Development of normalized consonant duration split by age and manner of articulation. Red squares indicate mean; black dots indicate outliners. Geminates are represented in green; Singletons are represented in pink. 183 Figure 20 Results of three-way interaction between manner of articulation (fricatives: red; laterals: blue; nasals: green; stops: purple), age (x- axis) and consonant category (left panel: Geminates, right panel: Singletons) with regards to the normalized consonant duration (y- axis). 184 Figure 21 Trilingual child: Development of normalized duration of Swiss German consonants split by age. Red squares indicate mean; black dots indicate outliners. Fortis stops are represented in green; Lenis stops are represented in pink. 185 Figure 22 Trilingual child: Development of normalized Swiss German consonant duration split by age. Red squares indicate mean; black dots indicate outliners. 186 Figure 23 Monolingual Italian control group: Development of normalized vowel duration split by age. Red squares indicate mean; black dots indicate outliners. Vowels preceding geminates are represented in green; Vowels preceding singletons are represented in pink. 187 Figure 24 Monolingual Italian control group: Development of normalized vowel duration split by age and vowel. Red squares indicate mean; black dots indicate outliners. Vowels preceding geminates are represented in green; Vowels preceding singletons are represented in pink. 188 Figure 25 Trilingual child: Development of normalized vowel duration split by age. Red squares indicate mean; black dots indicate outliners. Vowels preceding geminates are represented in green; Vowels preceding singletons are represented in pink. 189 XX | L i s t o f F i g u r e s Figure 26 Trilingual child: Development of normalized vowel duration split by age and vowel. Red squares indicate mean; black dots indicate outliners. Vowels preceding geminates are represented in green; Vowels preceding singletons are represented in pink. 190 Figure 27 Results of two-way interaction between consonant category and vowel and vowel and age. 191 Figure A Trilingual child: VOT development split by language over time; dashed line indicates the dividing line between short-lag and long- lag; red squares indicate the mean; fortis stops are represented in blue and lenis stops in pink iii Figure B Trilingual child: CD (normalized) development over time split by language and raddoppiamento fonosintattico (no = ITALIAN/yes = ITALIAN (RS)); red squares indicate mean; fortis stops are represented in blue and lenis stops in pink. iv Figure C Trilingual child: Geminates (pink) and singletons (green) split by manner of articulation. v Figure D Monolingual control group: Geminates (pink) and singletons (green) split by manner of articulation. v Figure E Trilingual child: Vowels preceding geminates (pink) and singletons (green) split by vowel type. vi Figure F Monolingual control group: Vowels preceding geminates (pink) and singletons (green) split by vowel type. vi L i s t o f T a b l e s | XXI LIST OF TABLES Table Caption Page Table 1 Overview on the monolingual control groups. 30 Table 2 The consonant phoneme inventory of German; Sounds in ( ) are allophones. 36 Table 3 The consonant phoneme inventory of Italian. 40 Table 4 The consonant phoneme inventory of Zurich German. 43 Table 5 The consonants shared between languages and consonants unique to one language. German only: orange, Swiss German only: red, Italian only: blue; shared between all: green, shared but with voicing deviance in Swiss German: light green, shared between Swiss German and German: pink, shared between Swiss German and German but with voicing deviance: light pink, shared between Swiss German and Italian: light blue, shared between German and Italian: grey. 47 Table 6 Overview of the grouped phonetic inventories. The asterisk indicates that phone occurred in 90% of the individual inventories. Phones without diacritic occurred in 50% of the individual inventories (Table modified after: Stoel-Gammon 1985, p. 506). 50 Table 7 Acquisition order of German phones (Table modified after: Fox 2007, p. 63). 52 Table 8 Acquisition order of the German phonemes (Table modified after: Fox 2007, p. 64) 53 Table 9 The development of the phonetic inventories of 13 children at age 1:6-2;3. The * indicates that 90% of the children produced the sound, the other phones occurred in at least 50% of the individual inventories (Table modified after: Zmarich & Bonifacio, 2004). 54 Table 10 Consonants present or absent in the consonant inventory of all children (Table modified after: Zanobini et al. 2012, p. 23) 55 Table 11 Percentages of consonants acquired (Table modified after: Zanobini et al. 2012). 56 XXII | L i s t o f T a b l e s Table 12 Comparison of the developmental steps in the acquisition of consonants; * indicates sounds that were produced by 90% of the children from the beginning. 58 Table 13 Overview of studies describing the bilingual acquisition of phoneme inventories. 61 Table 14 Analyzed material and percentage of disagreement in transcriptions between first and second transcriber. 72 Table 15 Monolingual German children: Individual and collective (≥ 50% of the children) overview of acquired, established and problematic sounds; * indicates that the sound is not acquired but only produced; () indicates sound which have been produced only once or in several token of the same word, i.e., not full filling the acquired criterion, but were always target-like 75 Table 16 Monolingual Italian children: Individual and collective (≥ 50% of the children) overview of acquired, established and problematic sounds; * indicates that the sound is not acquired but only produced; () indicates sound which have been produced only once or in several token of the same word, i.e., not full filling the acquired criterion, but were always target-like. 77 Table 17 Monolingual Swiss German children: Individual and collective (≥ 50% of the children) overview of acquired, established and problematic sounds; * indicates that the sound is not acquired but only produced; () indicates sound which have been produced only once or in several token of the same word, i.e., not full filling the acquired criterion, but were always target-like. 79 L i s t o f T a b l e s | XXIII Table 18 Trilingual child: Individual and collective (≥ 50% of the recordings) overview of acquired, established and problematic sounds; * indicates that the sound is not acquired but only produced; () indicates sounds which have been produced only once or in several tokens of the same word, i.e., not full filling the acquired criterion, but were always target-like. German results are displayed in the white left-most columns, Italian results are displayed in the grey left-most columns and the Swiss German results are displayed in the white right-most columns. 83 Table 19 Overview on contrastive features in Standard German, Zurich German and Italian stops. 111 Table 20 Overview of studies on VOT in monolingual first language acquisition. 113 Table 21 Overview of studies on VOT in multilingual first language acquisition. 115 Table 22 Monolingual Swiss German control group: VOT data; Numbers of fortis and lenis stops eligible for VOT analyses split by age. 127 Table 23 Trilingual child: VOT data: Left panel: Numbers of fortis and lenis stops eligible for VOT analyses split by age. Right panel: Numbers of prevoiced items across age and language. 128 Table 24 Trilingual child: VOT data: Prevoiced items in all three languages. 129 Table 25 Monolingual Swiss German control group: CD data: Numbers of fortis and lenis stops eligible for CD analyses split by age. 129 Table 26 Trilingual child: CD data: Numbers of fortis and lenis stops eligible for CD analyses split by age. 130 Table 27 Examples demonstrating the Swiss German vowel length contrast (Table modified after: Fleischer & Schmid 2006, p. 247). 160 Table 28 Comparison of the feature consonant and vowel length between Italian, Swiss German and Standard German. 161 Table 29 Studies on the acquisition of length contrast in consonants by monolingual children. 163 Table 30 Studies on length contrast in consonants in bilingual speakers. 166 XXIV | L i s t o f T a b l e s Table 31 Monolingual Italian control group: Number of analyzed consonants (geminates/singletons). 174 Table 32 Trilingual child: Number of analyzed consonants (geminates/singletons) of the trilingual child; *Numbers in brackets indicate the number of fortis sounds in the dataset. 175 Table 33 Trilingual child: Number of analyzed items (intervocalic Swiss German fortis/lenis stops). 176 Table 34 Monolingual Italian control group: Number of analyzed vowel (preceding geminates/singletons). 177 Table 35 Trilingual child: Number of analyzed vowels (preceding geminates/singletons). 178 Table A Exact MLU values, number of recordings included for MLU calculation and number of utterances evaluated for MLU calculation. vii 1 . I n t r o d u c t i o n | 1 CHAPTER 1 1. INTRODUCTION The topic of the present dissertation is the development of the sound system(s) of a trilingual (3L1) child, Tommaso1, acquiring (Standard) German2, Swiss German3 and Italian simultaneously. The overarching topic across all chapters is the longstanding question of language separation and cross-linguistic interaction (CLI) between the several first languages (L1s) of multilingual children. Specifically, the development of three segmental phenomena will be addressed: (1) consonant inventories, (2) stop patterns and (3) gemination. The phenomena will be analyzed based on a large corpus of spontaneous speech. The data was collected in a longitudinal study design and covers the time period of slightly more than a year (1;4 to 2;6) in all three languages. The analyzed age range provides insights into the early phase of development of the sound system(s) in 3L1 acquisition. In total, about 22 hours of trilingual speech have been analyzed. The present study would therefore be to date the most detailed study on phonetics and phonology in 3L1 acquisition. Trilingualism is an umbrella term for acquisition scenarios including three languages (Hoffmann, 2001b; Kupisch, 2023). It can refer to several speaker groups, for example, adults who learned two foreign languages (L2) besides their L1 or early bilinguals who acquired a foreign language. In this dissertation, trilingualism refers to the acquisition of three L1s simultaneously. Tommaso acquires the two home languages Swiss German and Italian in addition to Standard German, the language of the wider national society, simultaneously. This specific 3L1 acquisition setting falls into Hoffmann’s (2001b, p. 3) type (i) definition: “(i) [… c]hildren who are brought up with two home languages which are different from the language of the wider community.” While early bilingualism (2L1), i.e., speakers acquiring two L1s from early on, has been targeted in many studies and the body of literature is still growing (see Kehoe, 2015, for an overview on the development of phonetics and phonology in early 2L1 speakers), early 1 “Tommaso” is not the child’s real name. The name was chosen for data protection reasons. 2 I hereby mean the standard variety spoken in the south of Baden-Wuerttemberg which of course contains regional traits in all linguistic areas. Standard German and German are used interchangeably. 3 I will use Zurich German and Swiss German interchangeably. If another Swiss German variety is meant, it will be explicitly mentioned 2 | C h a p t e r 1 trilingualism is still under-researched. Empirical studies on the core areas of grammar in all three languages of the children are scarce (see Chevalier, 2011, Hoffmann, 2001b and Kupisch, 2023 for extensive overviews). To my knowledge, there are only three studies targeting the phonetic and phonological development of early simultaneous trilinguals (Mayr & Montanari, 2015; Montanari, 2011a; Yang & Hua, 2010). Results show, similar to 2L1 acquisition, evidence for separate language systems from early on and in some cases CLI (see Chapter 2), i.e., systematic interactions between the different L1s. However, there is an obvious difference between 2L1 and 3L1 speakers, namely the additional language in the acquisition context. The additional language adds a layer of complexity that should not be underestimated (Hoffmann, 2001a; Quay, 2008). The data collection and analysis are more time-consuming and methodologically challenging (Sṭavans & Hoffmann, 2015). Furthermore, homogenous groups of participants, i.e., speakers with the same language combinations and within the same age range, are more difficult to find. Despite these challenges and more importantly, the additional language makes 3L1 acquisition a prime candidate for disentangling potential predictors of CLI. The 3L1 acquisition of Swiss German, Standard German and Italian allows for shedding light upon questions concerning external CLI factors such as dominance, input quantity and quality (see Section 2.3.1.) as well as internal CLI factors such as structural ambiguity and markedness/complexity (see Section 2.3.2.). Furthermore, typological proximity (see Section 2.3.3.) between the languages might be an additional factor in causing CLI in 3L1 acquisition (Yang & Hua, 2010). This factor can be examined, as the child acquires two closely related Germanic languages (Standard German and Swiss German) and one Romance language (Italian). The sound system is an especially interesting area in 3L1 acquisition as in addition to the phonological and phonetic structures, the articulatory skills need to mature over time. For this maturation process, physical practice is needed. In 3L1 acquisition, possibilities for articulatory practice of language-specific phenomena are reduced in comparison to 2L1 and monolingual acquisition (under the condition multilinguals produce overall a comparable amount of speech output as monolingual and not three times as much). This could lead to the hypothesis that a trilingual’s sound system is especially vulnerable. As mentioned above, three phenomena (consonant inventories, stops and gemination) have been chosen as a focus for investigation in this thesis. These provide an interesting basis for CLI and language separation in 3L1 acquisition as described below. 1 . I n t r o d u c t i o n | 3 1. Consonant inventories (Chapter 4): The analysis of the consonant inventories allows one to compare the development of sounds that are (a) fully shared between the languages (e.g., /m/), (b) shared but are not congruent in all phonological or phonetic characteristics (e.g., /t/ (stop distinction depends on a different production pattern in each language, see also 2.) and (c) language-specific (e.g., /k͡x/ (Swiss German)). This is interesting for the following reasons: Firstly, the input quantity per language might play a role. The child needs to acquire 33 phonemes, i.e., ten to seven phonemes more than monolingual children (depending on the language), and to adjust eleven sounds to language-specific phonetic parameters (e.g., voicing is a feature of Italian and German /v/ while /v̥/ is unvoiced in Swiss German; stop patterns differ across languages (see 2.)). Secondly, language-specific sounds (quantitatively lower frequency of occurrence due to the relatively reduced input quantity) might be acquired later than shared sounds (quantitatively higher frequency of occurrence as they occur in all languages). Whether the shared sounds profit from a pooled input or not is not clear yet (see Section 2.3.2.3.). The shared sounds are in addition often the less marked and less complex sounds. Therefore, sharedness, markedness and complexity are difficult to tease apart. Thirdly, sounds that are shared but do not coincide in all (phonetic) characteristics might be subject to transfer or overgeneralization due to economy reasons. The phonetic differences need to be acquired and then to be mapped into the correct phonological categories and systems. These processes might be based on, e.g., typological proximity (Swiss German/German vs. Italian) or on the presence of the same parameter (e.g., voicing vs. devoicing (e.g., Italian/German: /v/ vs. Swiss German: /v̥/)) (see Section 2.3.3.). 2. Stop patterns (Chapter 5): Tommaso acquires three languages that all have binary stop distinction patterns. However, the basis of distinction between stop categories for each language is different. In Italian and German stops are distinguished based on Voice Onset Time (VOT). Italian fortis4 stops have short-lag VOT and lenis stops are prevoiced (true voice language), while German fortis stops have long-lag and lenis 4 For Italian and German, the terms voiceless and voiced are used interchangeably with fortis (voiceless) and lenis (voiced). Swiss German has no voicing contrast. Therefore, Swiss German sounds are exclusively labeled as fortis or lenis. For a more detailed discussion and explanation, see Chapter 5. 4 | C h a p t e r 1 stops short-lag VOT (aspirating language). All Swiss German stops have short- lag VOT. The contrast in Swiss German is based on closure duration (CD). Fortis stops have longer CD than lenis stops. Furthermore, the stops might be subject to CLI, as short-lag VOT is a parameter present in all three languages but it codes different categories in the languages. Therefore, the interpretation can be ambiguous due to structural ambiguity (see Section 2.3.2.2.) between the languages. In addition, the age of acquisition varies extremely between the languages. While the acquisition of the short-lag-long-lag distinction is in place around 2;0 (e.g., Snow, 1997), the acquisition in true voice languages takes up to the age of 7;8 (e.g., Khattab, 2007). The Swiss German CD contrast is under- researched and there is no reference age of acquisition. However, the CD contrast seems to be more difficult to perceive than the contrast in aspirating languages (Schönhuber et al., 2019) and therefore also suggests a late(r) acquisition in production. This might be especially true for the word-initial context and less so for a word-medial position, as e.g., long consonants including stops (word-medial) are acquired early (e.g., Vihman and Majorano, 2017; Kunnari et al., 2001). Given the factors just pointed out, the stop patterns are ideal for testing language separation for two reasons: The stop patterns differ in all three languages and the age of acquisition is different, too, as far as is known. Possible developmental scenarios can range from transferring one pattern (e.g., the less marked one) to the other language(s) (see Section 2.3.2.1), fusing similar stop patterns (e.g., based on topological proximity (genealogy; Swiss German/German vs. Italian) or having the same contrastive factor (VOT vs. CD)) (see Section 2.3.3.), or separate language systems. 3. Gemination (Chapter 6): In all three languages that Tommaso acquires duration is a distinctive factor. Consonant duration is distinctive in Italian and Swiss German but not in Standard German. Yet, consonant length is subject to different phonological and phonotactic constraints in Italian and Swiss German, respectively, and is therefore not equitable: In Italian, long consonants are distinctive in word-medial position only, while Swiss German allows for the distinction in almost all positions. Consonant duration in Italian is used for contrasting long (geminates, e.g., /pː/, /bː/) and short fortis and lenis consonants (singletons, e.g., /p/, /b/), whereas consonant duration in Swiss German is used for contrasting fortis (e.g., /p/) and 1 . I n t r o d u c t i o n | 5 lenis (e.g., /b̥/) sounds. This results in a four-way contrast in Italian (long vs. short in two voicing conditions), and a two-way contrast in Swiss German (fortis vs. lenis). This means that if input quantity (Section 2.3.1.1.) or structural ambiguity (Section 2.3.2.2.) are decisive cues in acquisition, the acquisition of gemination might be decelerated in 3L1 acquisition, possibly due to transfer or overgeneralization of the less complex and more economic Swiss German fortis- lenis contrast. Furthermore, gemination is a marked phenomenon. Therefore, also in this case, deceleration could be a possible CLI effect (Section 2.3.2.1.). Vowel duration is distinctive in both Germanic languages and a secondary cue in Italian. CLI into Italian could result in a more pronounced vowel duration contrast. Furthermore, irrespective of the respective phenomenon, dominance (Section 2.3.1.2.) can be a factor for CLI in the sense that features of the phenomenon in the dominant language influence the features in the weaker languages. However, as Tommaso is a balanced trilingual child (see Section 3.1.3), I will focus on the factor “balance” instead of “dominance” in my discussion. The findings from these empirical studies of the present dissertation will be discussed in the light of earlier research on CLI in multilingual first language acquisition and will provide evidence for language separation and monolingual-like development with little CLI. The current thesis is structured as follows. This introduction is followed by a background chapter on the differences in multilingual acquisition scenarios, the concept of CLI, potential predictors of CLI and a terminological note (Chapter 2). As the three empirical studies were all based on the same trilingual speaker and monolingual control groups, the participants as well as the data collection are described in Chapter 3 and will not be repeated in the subsequent chapters. The development of specific segmental phonological and phonetic phenomena will be discussed in detail in three empirical studies: consonant inventories (Chapter 4), stops (Chapter 5) and gemination (Chapter 6). Results of the three studies will be discussed and embedded in the larger context in Chapter 7 followed by a general conclusion presented in Chapter 8. 6 | C h a p t e r 1 2 . L a n g u a g e s e p a r a t i o n a n d c r o s s - l i n g u i s t i c i n f l u e n c e i n e a r l y m u l t i l i n g u a l s | 7 CHAPTER 2 2. LANGUAGE SEPARATION AND CROSS-LINGUISTIC INFLUENCE IN EARLY MULTILINGUALS 2.1. Differences between bilingual and trilingual language acquisition scenarios The most obvious and seemingly simple answer to the question as to what distinguishes 2L1 from 3L1 acquisition is the acquisition of three instead of two L1s. As Hoffmann (2001a) points out, this numerical difference between bi- and trilingual children results in quite a striking difference in terms of output scenarios: Trilinguals have seven different possible output scenarios and can choose to either speak all three L1s (ABC), one of three two-language-combinations (AB, AC, BC), or only one L1 (A, B, C). Bilinguals can choose from three options: speaking both L1s (AB), or only one of the L1s (A, B). A day has 24 hours, i.e., there is a natural limit for input and output quantity per day. The languages share this time. In a scenario with all three L1s, the sharing leads to an even more limited time for input per language, as the 24 hours are split between three languages instead of two. Analogously, the time for output per languages is more limited in 3L1 when compared to 2L1 acquisition. This is of course only true based on the assumption that trilinguals produce on average about as much output as monolinguals rather than three times as much, and that input givers do not provide three times as much input in trilingual as compared to monolingual situations. Furthermore, the third language brings an additional layer to code mixing/code switching5, i.e., there is additional source for these processes and three instead of two languages can be present in one utterance (Sivakumar, Müller, & Arnaus Gil, 2020). These differences in the acquisition scenarios could further lead to different degrees of balancing. Sṭavans and Hoffmann (2015) conclude in their overview paper that trilinguals are less balanced than bilinguals, although this does not imply that all bilingual children are balanced (see Section 2.3.1.2. for dominance as a predictor for CLI) or that trilinguals cannot be balanced (see Chapter 3.1.3). 5 Code mixing/switching refers to “the level of language choice [i.e.,] the use of language material from more than one language in discourse” (Sivakumar, Müller, & Arnaus Gil, 2020, p. 166). Sivakumar et al. (2020) use the code-mixing for unbalanced and code-switching for balanced multilinguals. 8 | C h a p t e r 2 In addition to the in- and output scenarios, the relation between language internal and external factors is more complex in 3L1 than in 2L1 acquisition. If two languages are acquired at the same time, the distribution of factors is more or less clear. An 2L1 example scenario could be that (ia) one language receives more input than the other, (iia) one of the languages is dominant, while the other is non-dominant, (iiia) a phenomenon is marked/complex in one language but not in the other, (iva) there is structural ambiguity or there is no structural ambiguity between the languages for certain phenomena. If three languages are acquired simultaneously, it might be more difficult to assign clear values to these factors. A possible scenario for 3L1 acquisition could be: (ib) one language receives the most input, one the second most and the third one the least input, (iib) likewise, the degree of dominance can vary across the languages (dominant, less dominant, weak), (iiib) a phenomenon can be marked/complex in e.g., one but not in the two other languages, (ivc) there can be structural ambiguity between e.g., all three languages. Sṭavans and Hoffmann (2015) point out that despite the above-mentioned differences, there are critical voices who argue that 3L1 acquisition is only a certain type of bilingualism. In my opinion, 3L1 and 2L1 acquisition scenarios are clearly different due to diverging input and output possibilities as well as a more complex relationship between language internal and external factors. I agree with Sṭavans and Hoffmann (2015, p. 147) who point out that “TFLA [trilingual first language acquisition] is not the sum of three first languages, nor the addition of a third language to bilingualism, but rather it is a unique phenomenon with its own characteristics and features that should be studied in its own right.” 2.2. Multilingual language acquisition and language separation Up until now, there is no framework or theoretical approach that is exclusively developed for research on 3L1 acquisition. 3L1 acquisition was mostly analyzed within the same theoretical frame as used in 2L1 acquisition. Therefore, the approach for analysis of 2L1 acquisition used in the present dissertation is reviewed in more detail in the next paragraphs. Systematic linguistic research on 2L1 language acquisition began in the 20th century (Leopold, 1947; Vogel, 1975; Volterra & Taeschner, 1978). One of the over- arching questions from early on was as to whether bilingual children have a single 2 . L a n g u a g e s e p a r a t i o n a n d c r o s s - l i n g u i s t i c i n f l u e n c e i n e a r l y m u l t i l i n g u a l s | 9 language system or several (separated) systems for their L1s. In simplified terms, the evolution of proposed answers to this question started with earlier concepts that understood a bilingual’s language system as a unified system of both languages (Volterra & Taeschner, 1978) and developed to an analysis of separated autonomous language systems (Meisel, 1989). Another approach for the analysis of bilingual speech was provided by Paradis and Genesee (1996). Paradis and Genesee (1996) proposed that bilingual children have separated language systems for their L1s that are interacting systematically. According to Paradis and Genesee (1996), this interaction can be grouped into three types: 1. Acceleration: A certain grammatical property emerges earlier in bilingual language acquisition than would be the norm in monolingual acquisition. 2. Deceleration6: A certain grammatical property emerges later in bilingual language acquisition than would be the norm in monolingual acquisition. 3. Transfer: A certain grammatical property is incorporated from one language into the other (where it is usually not present). These three types of interaction are subsumed under the term CLI. The first observations of CLI were made for syntactic phenomena. However, there is a high number of studies that have transferred the concept of CLI to phonetics and phonology as well (see Kehoe, 2015, for an overview7) and the large body of literature with research that is analyzing results in the light of the three types of CLI is still growing. Studies on the phonetic and phonological development of multilingual children confirmed that the phonological/phonetic systems are (mostly) separate and systematically interacting. Predictive and explanatory factors for CLI discussed in the literature are manifold and include the frequency of a phenomenon (Fabiano-Smith & Goldstein, 2010b), the complexity or markedness of a phenomenon (Kehoe, 2002), structural ambiguity between languages (Hulk & Müller, 2000; Paradis, 2001) and the language dominance of a bilingual speaker (e.g., Kehoe, Lleó, & Rakow, 2004; Paradis, 2001). The rich experience 6 The term used by Paradis and Genesee (1996) was “delay” and was linked to a slower rate of acquisition due to the burden of acquiring two languages simultaneously. Fabiano-Smith & Goldstein (2010b) proposed the term deceleration and I will follow this suggestion to prevent negative associations with the word delay. Further, I will follow the approach to see deceleration as the opposite of acceleration, i.e., not as the result of the burden of bilingual acquisition but in a more neutral way (see Kehoe, 2015, for a discussion on the term). 7 To my knowledge, Kehoe (2015) is the most extensive and accurate overview and discussion on CLI in phonetic and phonology. Therefore, it will be referred to quite often in Chapter 2. 10 | C h a p t e r 2 and evidence from studies in 2L1 acquisition are a good starting point for the investigation of 3L1 acquisition. 2.3. Predictors of Cross-linguistic Interaction (CLI) During past years of research on multilingual L1 acquisition, several predictors and explanatory factors potentially accounting for CLI have been discussed. This section presents the most discussed factors for CLI in phonetics and phonology. First, language external factors: input (quantity, connected with frequency of language use) and quality, as well as metalinguistic awareness, see Section 2.3.1.1.) and dominance (see Section 2.3.1.2.) and second language internal factors: complexity/markedness (see Section 2.3.2.1.), structural ambiguity (see Section 2.3.2.2.) and frequency (of occurrence, see Section 2.3.2.3.) will be reviewed. Further, the factor typological proximity (see Section 2.3.3) will be discussed. Typological proximity is new to the set of CLI predictors as it can be only investigated directly in 3L1 acquisition. As pointed out already above, 3L1 acquisition has not received as much attention as 2L1 acquisition. The following subsections will start with evidence from studies on 2L1 acquisition if 3L1 studies are not available. Those 2L1 studies are concerned exclusively with phonetics and phonology. Additionally, aspects used as explanatory factors for CLI within 3L1 acquisition will be presented to complete the picture. As the literature on 3L1 phonetics and phonology is still limited, studies on other linguistic core areas have been integrated as well. 2.3.1. External factors 2.3.1.1. Input Input as a super-category unites two related CLI predictors: quantity and quality of input. Quantity of input is difficult to capture but by large it is determined by the number of hours during which children hear a language. There are tools such as Language ENvironment Analysis (LENA), which estimate input (number of words) (Orena et al., 2019) but such detailed data is rather the exception than the norm in most studies. As already suggested in Section 2.1., the relatively reduced quantity of input might play a role for the rate of acquisition in 3L1 acquisition (e.g., Lleó, 2016). However, the language with the highest amount of input is not necessarily the one developing the fastest. Yang and Hua (2010) studied the phonological development of a Taiwanese- 2 . L a n g u a g e s e p a r a t i o n a n d c r o s s - l i n g u i s t i c i n f l u e n c e i n e a r l y m u l t i l i n g u a l s | 11 Mandarin-Spanish trilingual child. Mandarin and Spanish each made 40% of the input, while the share of Taiwanese was only 20%. The Spanish sound inventory developed fastest in accordance with the assumption that more input leads to faster acquisition. In contrast, this trend was not mirrored for Mandarin. Despite the lower proportion of input, the Taiwanese sound system appeared to develop faster than the Mandarin one. In other core areas, quantity of input (or frequency of language use) was also found to have an effect in 3L1 acquisition. Montanari (2010) found an effect in the context of vocabulary acquisition. The author studied the cumulative vocabulary of a Tagalog- English-Spanish trilingual child. She analyzed the extent of phonetically distinct doublets (concepts with translation equivalents in two languages) and triplets (concepts with translation equivalents in two languages) in the child’s lexicon as well as neutral words (words existing in all languages such as mommy/mommy/mami ‘mom’). The results showed that neutral terms made up the highest share of the lexicon. However, signs of language separation were present from early on. Tagalog had the highest quantitative input rate and seemed to be the dominant language. This might account for the fact that at an initial stage one of the translation equivalents was from Tagalog and one from English or Spanish. Triplets which had English and Spanish equivalent pairs did not appear before 1;7. This means that the child first acquired the word in her dominant language and one of her weaker languages, suggesting that not only neutral words existing with a different form in all languages were important but also especially words that allowed the child to interact in Tagalog the language most frequently used. Besides reduced quantity, qualitative input factors such as accented speech (e.g., Lleó, 2016), input contexts, i.e., where and with whom the language is spoken (Quay, 2011), and discourse style of the interlocutors (Chevalier, 2013) might play a role as well. The analysis of qualitative input needs to be tackled from various perspectives and ideally combines the number of input givers and their proficiency. Mayr and Montanari (2015) analyzed stop productions of two English-Italian- Spanish trilingual sisters. While the English and Spanish (voiceless) stops did not show any diverging pattern from adult targets, the Italian stops did. The father and other native speakers in the sisters’ environment (English was the majority language) provided English input; their Spanish input was provided exclusively by their Mexican nanny. The Italian input, in contrast, was provided by their native Italian mother but also shaped by the English-accented input from their English-dominant peers in a bilingual school 12 | C h a p t e r 2 program. The authors hypothesize that CLI in the Italian productions might be the result of input quality. Mayr and Montanari (2015) conclude that a single input source might be beneficial for phonological acquisition and prevent CLI, as the values for the Spanish stops were very close to the stops provided in the input by the Spanish-speaking nanny. Quay (2011) studied the language development of two trilingual children; a trilingual Chinese-English-Japanese girl and a German-English-Japanese boy. She compared the home languages with the language spoken at the kindergarten (Japanese). Both children used predominantly Japanese at the kindergarten, i.e., showed a clear separation between day care and home languages and were sensitive to the monolingual setting. The girl produced fewer utterances in the recordings in the daycare setting than in the home setting. Nevertheless, “Japanese was [her] strongest language even though she had fewer opportunities to speak in multi-party conditions at daycare than in dyadic conditions at home” (Quay, 2011, p. 39). Besides the input of the caregivers, input from peers was helpful as well. This support is not correlated with quantity of input but rather with “[…] their repetitions of adult and child speech and their modelling of language and behavior at a level shared by other child participants” (Quay, 2011, p. 39). Quay (2011) argued that a combination of quantitative and qualitative factors helps to acquire languages. Chevalier (2013) argued that the reaction of the interlocutor had an effect on the active trilingualism in terms of language mixing. She compared the reactions of interlocutors to language mixing of a Swiss-German-French-English child in the home languages French (provided by the father) and English (provided by the aunt). While the aunt repeated the non-target language words in the target language in most cases, i.e., “repaired” the utterance, the father “ignored” the language mixing and moved on within the conversation. These strategies led to less mixing in English than in French despite the fact that there was considerably less quantity of input for English. However, besides the quantity of input and the reaction of the adults to language mixing of the child, other social factors played a role as Chevalier (2013) hypothesized. The degree of closeness of the relationship and which role the input provider took on were important as well. While the aunt “[had taken] on a teaching role” (Chevalier, 2013, p. 29) the parent-child- relationship had other priorities than language teaching. An effect related to both input components in a wider understanding is metalinguistic awareness. Bosch and Sebastián-Gallés (2001, p. 79) suggested that a multilingual setting can result in “enhanced language discrimination capacity” as the 2 . L a n g u a g e s e p a r a t i o n a n d c r o s s - l i n g u i s t i c i n f l u e n c e i n e a r l y m u l t i l i n g u a l s | 13 linguistic surrounding of children acquiring more than one language facilitates uncovering language-specific and cross-linguistically diverging sounds and patterns as they receive input in several languages, i.e., they understand the phenomena on a meta level and are more aware of differences. The term does not seem to be directly correlated with quantity or quality of input but rather with the fact that multilingual children receive input in several languages. In a study on the consonant inventories of the abovementioned child acquiring English, Spanish and Tagalog, Montanari (2011) raised the question whether trilingual children’s phonological acquisition benefits from their more advanced discrimination abilities and their higher attention towards different sound systems. She claimed that the better perceptive abilities of multilinguals result in a boost of their productive abilities (Montanari, 2011). The child showed early language separation and accelerated acquisition of some segments compared to monolingual children. Within the same line of argumentation, Montanari (2009a) explained the results obtained in her study on early multi-word combinations of the same child. In the study on syntactic development, the child showed language separation from early on and produced different word orders that were target-like for the respective language. The author suggests that increased input diversity might result in an enhanced language discrimination capacity. Consequently, the child paid more attention to the different structural properties of the L1s. This higher attention positively influenced language separation. To sum up, the effect of input and whether quantity or quality is crucial is not yet clear. The interaction is complex and both factors are important. Input is indispensably linked with language acquisition and all CLI predictors. If there is no input in another language, children are not multilingual, as input is the source for language acquisition. Input itself is a complex construct and the totality of all other CLI predictors. Therefore, input is a decisive factor for CLI and needs to be taken into consideration. 2.3.1.2. Language dominance Language dominance has often been used as predictor or as explanatory factor for the direction of CLI. The dominant or stronger language of multilingual speakers is usually the one that they hear and use the most (Kehoe, 2015). Especially from the moment they enter school onwards, the dominant language is often the majority language (the language of the wider national society they live in) and consequently typically their most proficient language. If dominance is defined by proficiency, the majority language is not necessarily the dominant during every period of language acquisition. Especially during the first 14 | C h a p t e r 2 years, before attending kindergarten or similar institutions, the dominant language might be the language spoken at home or by the caregiver the children interact the most with. This language can also be a language that is not that of the wider national society. In several studies on the phonology of bilingual speakers, the dominant language appeared to have an influence on the minority language. Kehoe et al. (2004) investigated VOT acquisition in German-Spanish bilinguals. German has a long-lag/short-lag-VOT contrast, while Spanish contrasts between prevoiced stops with negative-lag and short- lag stops. The authors explained the transfer of long-lag VOT from German into Spanish with the German-dominance of a child. However, transfer was only found for one child in the study by Kehoe et al. (2004). Similarly, Paradis (2001) suggested dominance as an account for CLI from French found in English truncation patterns of French-English bilinguals (see Section 2.3.2.1. for another possible explanation of CLI in this study). However, there are also studies showing bidirectional CLI. For example, Stoehr et al. (2018) showed influence form both languages in the VOT patterns of Dutch-German bilingual children. While the Dutch voiceless stops had target-like values (short-lag VOT) the VOT in German was not produced target-like. The voiced stops were neither target- like in Dutch nor in German. An exclusive source of influence could not be identified. Almeida et al. (2012) investigated branching onsets and word-medial codas in a French- Portuguese child. The authors reported French influence in the Portuguese productions and Portuguese influence in the French productions. Consequently, they concluded that dominance alone could not be the only explanatory factor for the CLI found, as a child cannot be dominant in two languages at the same time. In studies on the syntax and lexicon of trilingual children, the weaker language was sometimes found to have an impact. Devlin et al. (2015) analyzed the productions of an English-Italian-Scottish Gaelic trilingual child. The child’s dominant language was English. The child (recordings from age 2;3-3;7) produced clitic-right dislocation (involving clitic-doubling) as well as subject dislocations to the right and left in English, e.g., “He’s give it back the muffin” (Devlin et al., 2015, p. 104). Clitic-right dislocation is ungrammatical in English but grammatical in Italian. Contrarily, subject dislocations are an English linguistic property; however, this construction is usually not yet acquired in English at the investigated age in monolingual acquisition. The authors argued for CLI from the non-dominant language Italian into English as dislocations were frequent in Italian. 2 . L a n g u a g e s e p a r a t i o n a n d c r o s s - l i n g u i s t i c i n f l u e n c e i n e a r l y m u l t i l i n g u a l s | 15 Devlin et al. (2017) also found CLI from the non-dominant language into the dominant language in the English-Italian-Scottish Gaelic trilingual child just mentioned. Their study on the unaccusative/unergative distinction in English and Italian was based on grammaticality judgements. While in Italian the distinction is based on auxiliary selection (unaccusative: Il treno è arrivato. direct translation: ‘The train is arrived’ lit: ‘The train arrived’; unergative: Maria ha cantato, direct translation: ‘Mary has sung’ lit: ‘Mary sang’), the distinction is not overly marked by auxiliaries in English. The results showed that the child acquired the distinction in both languages. However, the percentage of correctness was higher for the English (perfect participle task: unaccusative correct: 75%; unergative correct: 75%; resultative grammatical judgment task: unaccusative correct: 75%; unergative correct: 91%) compared to the Italian judgements (auxiliary selection: unaccusative correct: 58%; unergative correct: 75%). There was an overlap of problematic verbs in Italian and English. Therefore, the authors argued that the child’s development in the stronger language English was decelerated by the weaker language Italian. Also Kazzazi (2011) found influence from the “lesser-dominant” language Farsi in productions of two German-English-Farsi trilingual siblings. German was the children’s dominant and English their weakest language. Farsi influence in German and English became evident in post-modification in compounds (e.g., Feuerlager instead of Lagerfeuer ‘bonfire’), possessive pronouns and the use of light verb constructions. Furthermore, the children used the Farsi ezafe morpheme (“a kind of ‘default’ relating particle”, Kazzazi, 2011, p. 67) in German and English. Kazzazi (2011) hypothesized that it was not dominance but a cognitive tendency towards transparency and uniformity that played a decisive role in triggering CLI. This cognitive tendency was satisfied by Farsi. CLI occurs bi-directional also in 2L1 children and furthermore dominance is not the only driving force of CLI.8 In summary, dominance seems to be a factor that influences type and direction of CLI. However, it cannot be the only driving force of CLI, as effects of the minority language can be found in the dominant language and bidirectional CLI occurs as well. 8 The author further provided new terms for the direction of CLI in 3L1 acquisition: ‘forward influence’ for CLI from the strongest into both non-dominant languages, ‘reverse influence’ from the weakest into the dominant language and ‘lateral influence’ between the non-dominant languages. Whether this new terminology was required is debatable, from my point of view the already existent terms can describe those processes as well. 16 | C h a p t e r 2 2.3.2. Internal factors 2.3.2.1. Complexity or markedness of a phenomenon The definition of complexity used in the present dissertation is the following: “In general, a phonetic/phonological property that contains more elements (e.g., features), more structure, or is more difficult to produce is more complex than a phonetic/phonological property that contains fewer elements, less structure, or is less difficult to produce” (Kehoe, 2015, p. 151). Following this definition affricates should be more complex than stops or fricatives on their own. Markedness means the occurrence and frequency of a certain phonetic/phonological phenomena in the languages of the world (Jakobson, 1941). As Kehoe (2015) points out, the definitions behind complexity and markedness are difficult to capture, as they can mean the same or different concepts: The two concepts are often correlated as complex properties are often more marked than simpler properties. In this dissertation, complexity and markedness will be treated as separate factors in the analysis. This distinction is not made in all studies; sometimes complexity and markedness are used as quasi-synonyms (Kehoe, 2015). It is usually assumed that unmarked and less complex phenomena are acquired before marked and complex phenomena (e.g., Jakobson, 1941; Kehoe, 2015). With regards to types of CLI, studies on marked/complex phenomena show mixed results. While the acquisition of vowel length, a rather marked phenomenon, was decelerated in German-Spanish bilinguals (Kehoe, 2002), other studies showed acceleration effects for marked phenomena such as Tamburelli, Marco, Sanoudaki, Eirini et al. (2015) for onset clusters, in Polish-English bilinguals. Kehoe (2015) reflected on how to account for these different types of CLI triggered by the same predictor. She stated that a possible account might be whether a phenomenon is present in only one or in both L1s. Vowel length is only present in German, while onset clusters exist in Polish and English9. This can lead to the assumption that marked/complex phenomena only present in one language might be prone to deceleration, while a marked/complex phenomenon present in both languages is rather accelerated. This approach was also suggested by Lleó (2002). However, Almeida et al. (2012) studied branching onsets and rhymes in a French-Portuguese bilingual child (see 2.3.1.2.). They found acceleration in onsets but not in codas despite the fact that both phenomena are present in both languages. This means that the presence 9 Clusters in Polish and English are subject to different constraints. However, the English clusters are a subset of the Polish clusters as Tamburelli et al. (2015) point out. 2 . L a n g u a g e s e p a r a t i o n a n d c r o s s - l i n g u i s t i c i n f l u e n c e i n e a r l y m u l t i l i n g u a l s | 17 of a phenomenon in both languages alone is not enough to account for types of CLI. The authors hypothesized that other factors, in this specific case “structural ambiguity” (see Section 2.3.2.2.), need to be taken into consideration as well. To conclude, complexity and markedness play an important role in monolingual acquisition (e.g., Jakobson, 1941). In multilingual acquisition, the analyses of marked and complex phenomena yield mixed results and cannot explain the order/process of acquisition alone. 2.3.2.2. Structural ambiguity The concept behind structural ambiguity is based on the assumption that CLI occurs at interfaces between grammar modules of the bilingual’s two L1s if there is structural overlap at the surface level. This overlap can lead to ambiguous interpretations of the respective phenomena for bilingual speakers (Müller & Hulk, 2000). The predictor structural ambiguity has primarily been used for explaining CLI at interfaces with syntax (e.g., Hulk & Müller, 2000). This approach sees interfaces as especially vulnerable, e.g., the syntax-pragmatics interface but also at interfaces with phonetics and phonology (e.g., question intonation) or between phonemes and phones (e.g., short-lag VOT, Lleó, 2016). An example for structural ambiguity is short-lag VOT. Short-lag VOT overlaps on the surface and is at the interface between phonetics and phonology as it denotes, e.g., lenis sounds in German but fortis sounds in Italian. The phonetic measure (VOT in ms) needs to be translated to phonology (categories). The interpretation and categorization of short- lag can therefore be ambiguous for German-Italian multilinguals. Only a few results of the studies in phonetics and phonology were explained with structural ambiguity, e.g., Paradis (2001) and Almeida et al. (2012) (see Section 2.3.1.2. (dominance) and Section 2.3.2.1. (compelxitiy/markedness)). Paradis (2001) used structural ambiguity as explanation for CLI in truncation patterns, i.e., the omission of syllables, produced by French-English bilinguals. She suggested that there was a potential structural overlap between two specific word rhythm patterns in French (WWWS)10 and English (WS’WS). Only truncations for English words with this specific rhythm pattern differed from truncations of the monolingual control group. Paradis (2001) explained this with CLI from French into English, as WWWS was the only French word rhythm type. Almeida et al. (2012) proposed that structural ambiguity might explain CLI in cluster 10 S = strong, W = weak 18 | C h a p t e r 2 productions of French-Portuguese bilinguals rather than the markedness/complexity, which was the predictor more closely examined in the study. Lleó (2016) is not in favor of a structural ambiguity-based explanation for CLI in the sound systems. She provided an overview on several phenomena at the interface with phonetics or phonology and at the interface between segmental (13 phenomena) and supra-segmental (7 phenomena) components in Spanish-German bilinguals. Various types of CLI (the reference points are the Spanish phenomena) can be found in these constellations (acceleration, no effect, short delay, variation/delay, delay, delay with transfer, long delay with transfer, Lleó, 2016, p. 353). Relevant phenomena discussed by Lleó (2016, p. 353) for the present dissertation are voiced stops (acceleration), short VOT (variation/delay), pre-voicing (long delay with transfer), and spirants (long delay with transfer). However, Lleó (2016, p. 544) concluded that the “[a]ttempts to predict and explain the outcomes of language contact go from internal and/or external factors to the interfaces, findings still being incomplete. It is shown that in the phonological domain special attention should be devoted to internal factors like markedness, uniformity and frequency.” Overall, it is unclear whether structural ambiguity is a factor for CLI in the sounds systems because studies explaining their findings with structural ambiguity are scarce. 2.3.2.3. Frequency (of occurrence) Frequency is the (degree of the) presence or absence of segmental or supra-segmental phenomena in a language (Kehoe, 2015). It has been shown for monolinguals that “young children are sensitive to the statistical properties of the ambient language” (Kirk & Demuth, 2003, p. 447), meaning that phenomena with a higher frequency are acquired before less frequent phenomena. According to Kehoe (2015) the concept and consequences of frequency are difficult to capture in multilingual language acquisition. It is not clear whether there is “pooling”, i.e., a kind of average of a phenomenon’s frequency across the L1s that might lead to acceleration (Kehoe, 2015); or if the higher frequency of a phenomenon in one L1 alone is enough to trigger acceleration, i.e., ‘positive transfer’ as suggested by Goldstein and Bunta (2012) amongst others. It is difficult to disentangle the two kinds of frequency effects (Kehoe, 2015). Based on a study by Kehoe and Lleó (2003) on the acquisition of onset and coda clusters in Spanish- German bilinguals, Kehoe (2015) reasoned that the idea of ‘positive transfer’ seems slightly advantaged over pooled input: Coda clusters are frequent in German but 2 . L a n g u a g e s e p a r a t i o n a n d c r o s s - l i n g u i s t i c i n f l u e n c e i n e a r l y m u l t i l i n g u a l s | 19 extremely infrequent in Spanish. In contrast, onset clusters are present in both languages. In a pooled input scenario, onset clusters would be more frequent than coda clusters. Yet, coda clusters were acquired before onset clusters by the bilinguals, similar to the order of acquisition in monolingual children (Kehoe & Lleó, 2003). This suggests that pooled input frequency does not affect the order of acquisition as higher frequencies due to the presence of a phenomena in both languages did not per se lead to an effect at all (Kehoe, 2015). Findings regarding frequency effects in 2L1 language acquisition are mixed as acceleration (Keffala, Barlow, & Rose, 2018) and deceleration (Gildersleeve-Neumann, Kester, Davis, & Peña, 2008) were reported for the same phenomenon (coda acquisition) within the same population (English-Spanish bilinguals). Furthermore, other studies such as Fabiano-Smith and Goldstein’s (2010b) investigation of phonetic inventories of English-Spanish children did not report any effect of frequency on the accuracy of shared sounds. Tamburelli, Marco, Sanoudaki, Eirini et al. (2015) also explained that frequency was not the decisive factor in their study on the acquisition of clusters in English-Polish bilinguals. These findings suggest that pooled input (or at least a positive effect of it) is unlikely. Overall, frequency (of occurrence) does not seem to be a reliable predictor or explanatory factor for CLI in bilingual language acquisition as the results show either opposite results or no effects. Furthermore, another type of frequency, i.e., frequency of language use, and amount of input are not differentiable. If language A provides more input than the other(s), the language-specific phenomena of language A are very likely to have a higher frequency. The results obtained by Montanari (2010) on the lexicon (Tagalog being the dominant language with the most input and therefore the one in which new words were acquired first) could also be explained with a higher frequency in the input as Tagalog provided the most input (see Section 2.3.1.1.). 2.3.3. A “new” factor: Typological proximity Typological proximity is typically seen as genealogy (i.e., language families such as Germanic languages or Romance languages) or in terms of perceived distance between the languages by the learner (psychotypology, Kellerman, 1983). Liceras and de la Fuente (2015) make a similar distinction for L2 acquisition contexts in a study on the interlanguage of French L1 speakers acquiring Spanish as L2. They diff