• Keine Ergebnisse gefunden

3.1 Third language acquisition versus second language acquisition

3.1.3 Evaluation

The preceding discussion of the heterogeneous studies and results clearly shows how lively the debate about cross-linguistic influence in the field of language acquisition still is. We do not promise to find the perfect answer with the current study, but we hope to add another aspect to this debate, which could answer some of the questions raised and could provide further evidence in support of available models for third language acquisition.

Before we describe the layout of this study (in Chapter 5), we will offer an explanation for the, at first sight, contradicting results of the previous studies. First, let us have a brief look

40 at one article that offers a systematic review of 71 studies focusing on transfer in L3 acquisition (Puig-Mayenco et al. 2018), which is of course much more than what was covered in the chapter above. Puig-Mayenco et al. (2018: 18) prominently demonstrate that all studies they included in their meta-analysis revealed a significant amount of variation. Most importantly, we find variation across all domains “that is, differences exist related to the backgrounds of the subjects tested, the languages in the trilingual parings, the domains of grammar tested and several non-trivial distinctions in type, creation and administration of the testing methodology” (Puig-Mayenco et al. 2018: 18). This is their main argument for why there are so many different models that all try to capture transfer in L3 acquisition. We follow a different evaluative strategy, but many points that we will present in the next paragraphs find support in Puig-Mayenco et al. (2018).

The most prominent factor is perhaps the sample size of the studies – they differ crucially and are typically relatively small. Compare for instance Bardel and Falk (2007) who looked at nine participants, Håkansson et al. (2002) whose study contained 20 subjects, or Lloyd-Smith et al. (2018) who analyzed 21 third language learners. We assume that these low numbers were chosen to create a (more or less) homogeneous group. In Bardel and Falk, for instance, we find that “[t]he learning was the same for all participants” (2007: 470). The participants received the same language instruction and had the same input, which makes a comparison possible. Yet, in total, they analyzed only nine L3 learners. Therefore, the findings hardly qualify for any generalizations.

Furthermore, in most studies, only one grammatical phenomenon or area was included in the analysis (compare Flynn et al. 2004: relative clauses). Again, this does also not allow a researcher to formulate generalizations for the acquisition of a language in general. Siemund et al. (2018) mention this typical shortcoming of most studies that investigate cross-linguistic influence in L3 acquisition: “the examination of one particular phenomenon is usually taken to be sufficient to allow for far-reaching generalizations” (2018: 384). It is of course not possible to analyze all aspects of a language within the limitations of one study, and we will also not be able to do this here in this study. Yet, this problem of only analyzing one feature is especially relevant since some of the previous research findings point towards property-by-property transfer (Westergaard et al. 2017). How could claims be made if we only looked at one area, such as adverbial placement, demonstratives, or subject-verb-agreement? In addition, Lorenz et al. (2018) demonstrate that frequency of occurrence in the respective language affects cross-linguistic influence. Slabakova (2017: 662) also argues that one additional factor that influences

41 the acquisition of a L3 is the “construction frequency in the target L3”. This needs to be included in further research.

Furthermore, as was likewise stressed by Hopp et al. (2018) and Hopp (2019), most of the studies exclusively analyzed adult language acquisition but paid little attention to child L3 acquisition. Hence, the groups of language learners that were analyzed are essentially different types of learners, and therefore, it may not be advisable to compare child language acquisition with adult language acquisition. Chapter 3.4 will further investigate why we should differentiate between adult and child language acquisition and why it is therefore relevant to conduct the current study.

However, not only child versus adult language acquisition was explained to have an influence in the process of acquiring a foreign language and the performance in that particular language, but also the status of the languages and level of proficiency in the previously known languages was mentioned to have a significant influence. The status of the languages concerns the debate whether bilingual participants are equally proficient in both languages, such as balanced bilinguals, or whether they have one majority language and one minority language and would then be considered unbalanced bilinguals. Furthermore, the status of the L2 could be different from speaker to speaker. The L2 could be strictly speaking a second native language (if a person grows up in a bilingual community, for instance), it could be a foreign language acquired via formal instruction in school (such as English acquisition in Germany by monolingual German speakers), the L2 could be acquired during childhood or as an adult, or the L2 could be the majority language of the country of immigration. The latter would result in bilingual heritage speakers that have knowledge of a heritage language (the majority language of their former country of residence or that of their parents or grandparents) and that acquire at some point in time, after immigrating to a new country, the official language of the new country of residence as a L2 (an example would be a Russian speaker that moves to Germany). More about heritage speakers will be explained in Chapter 3.5.

Lorenz and Siemund (forthc.) and Lloyd-Smith et al. (2018) argue that most models that explain the role of previously acquired languages in L3 acquisition are based on participants who grew up in a monolingual setting with one language, who then received formal education in a L2, and acquire their L3 mostly in a university setting. To name just the most prominent models, we find this situation in Na Ranong and Leung (2009), Bardel and Falk (2007), Flynn et al. (2004), and Rothman (2011). However, when we look at the more recent studies that were previously discussed, and when we consider a development that Kupisch et al. (2013) describe, we notice that there is a shift towards analyzing a different type of L3 learner (Lorenz &

42 Siemund forthc.). We find young bilingual heritage speaker who grow up with a heritage language and a majority language. Heritage speaker grow up bilingually, usually either as simultaneous bilinguals or early, sequential bilinguals, with a minority language, their family language or heritage language, and with a majority language, the community language of the country of residence (Montrul 2016: 2). An example of such a speaker would be a child of parents that grew up in Turkey and speak Turkish as their native language and who immigrate to Germany when the child is still very young. They continue to speak Turkish at home, but the child is also exposed to German outside of their home and in school. We will discuss the concept of heritage speakers in more detail in Chapter 3.5, but for now, we can notice that there is a difference between this type of bilingual speaker and a bilingual speaker that receives formal education in their L2.

Keeping this in mind, it seems less surprising that learning a third language may follow different patterns depending on the status of the previously acquired languages. We therefore see the need for clearly assessing the type of bilingual speaker when analyzing cross-linguistic influence in L3 acquisition. As research that focused on such heritage speakers has shown, it seems impossible to replicate findings of former studies that analyzed different types of L3 learners (see for example Westergaard et al. 2017; Hopp 2019; Siemund et al. 2018). Hopp emphasizes that the status of the L1 and the L2 in heritage speakers is not straightforward: “the heritage language may – strictly speaking – not be the L1 of these children, and the early-acquired other language (German) may equally be a L1 or may have taken over the role of the L1 as it became the more dominant language” (2019: 579). Dewaele (1998: 29) also stressed

“that the L1 is not necessarily always the dominant active language.” Therefore, models that propose exclusively transfer from the L1 or the L2 may be difficult to apply in such heritage speaker situations. It seems as if the status of each language, i.e. which language is the dominant or majority language and which language is the minority language, affects the learning process of the L3. As was briefly reported, Lloyd-Smith et al. (2017) were able to demonstrate in an accent-rating study that dominance patterns of the L1 and L2 of heritage speakers determine CLI in the L3. Their results suggest that in the L3 English, at least in terms of foreign accent, there are differences between those that were more or less proficient in the heritage language Turkish: transfer seemed to come predominantly from German (the overall dominant language) but they identified that this was different for those who were more proficient in their L1 Turkish (Lloyd-Smith et al. 2017: 156-158). They showed that both the L1 and the L2 could be sources of CLI in L3 acquisition for heritage speakers and that proficiency in the heritage language affects the outcome in pronunciation. In another study, however, Lloyd-Smith et al. (2018),

43 these results could not be replicated. Here, when they analyzed syntax, the proficiency level of the heritage language had no impact on the results; all participants showed CLI from both the dominant language German and the heritage language Italian (Lloyd-Smith et al. 2018: 156) which ultimately supports Westergaard et al. (2017) and Slabakova (2017). Once more we find an indication that more research is needed, in order to understand how CLI of the L1 and the L2 influence the L3 (especially in unbalanced bilingual heritage speakers).

This argument about language dominance, i.e. balanced versus unbalanced bilinguals, finds little attention in Puig-Mayenco et al.’s (2018) comprehensive review of L3 acquisition studies. Yet, we are convinced that the type of bilingual learners that are investigated crucially affect the role that the previously acquired languages play in additional language acquisition.

In addition, the status of the L3 is also important; differences are to be expected concerning the initial, intermediate, or advanced stages of the third language. Several former studies have explicitly limited their findings to the initial stages/initial state8 of the L3 acquisition process (Hermas 2014; Fallah & Jabbari 2018; Rothman 2011). Slabakova (2017) also proposes that L3 studies should go beyond the initial stages of learning an L3 but to also focus on the developmental processes. This developmental perspective may offer even further insights into understanding how cross-linguistic influence determines the acquisition of a third language. Lorenz et al. (2018: 10) have shown, on the basis of a cross-sectional study, that there were hardly any differences between the different groups in the older cohort; here, both mono- and bilingual learners performed comparably. Yet, among the younger learners, Lorenz et al.

(2018: 10) identified differences between L2 and L3 learners. This clearly shows that further studies should specify the current status of the L3 and that we need to put more emphasis on cross-sectional studies that include several proficiency levels or, ideally, that there are more longitudinal studies that follow a number of students over a longer period of time.

So far, as was shown, many studies concentrated on adults rather than children. The groups were mostly homogeneous but presented therefore only one particular type of L3

8 Several authors refer to either the initial state or the initial stages, hence, it seems as if both concepts are sometimes used synonymously. Take Rothman (2011), for instance; here, he limits the applicability of the TPM to the initial state (Rothman 2011: 112). Later, in Rothman (2015) he acknowledges that he is actually not referring to the initial state in accordance to Schwartz and Sprouse (1996), but that he should use the term initial stages instead. Initial state, following Schwartz and Sprouse (1996: 41), is the starting point in L2 acquisition, namely the entire grammar of the L1, at the onset of L2 acquisition. In L3 acquisition, the initial state would then be both grammatical systems of L1 and L2, because these two systems are theoretically available in further language acquisition (Rothman 2015: 179). What Rothman (2011) and arguably many others are actually referring to are the initial stages. Initial stages could be defined as “the period in which structurally driven wholesale transfer from the L1 or the L2 takes place” (González Alonso & Rothman 2017: 688). Hence, it is the time between first initial exposure and the time when the learner of the L3 has had already some (limited) learning experience of the L3 (Westergaard et al. 2017: 669).

44 learner. Only some studies looked at younger learners and included bilingual heritage speakers, which are an increasing group of people in our societies (see the introductory chapter). In the current study, it is aimed to address some of the aforementioned deficiencies; however, new deficiencies are the consequence (see Chapter 8). The sample size will be larger, we will not focus on 10 speakers exclusively, but the entire sample consists of 249 speakers. This increases the likelihood for proposing adequate generalizations. In addition, the groups that will be looked at comprise multiple subgroups, intermediate and advanced learners of English for both monolingual and bilingual children. This allows for a wider picture since various language combinations are included. However, this means that the individual subgroups are also comparably small. Concerning the grammatical area that will be analyzed, we are also limited to one specific area as we can hardly make a general statement about language acquisition regarding all fields of grammar within one study. It is not possible to cover every grammatical area, from phonetics and phonology, to the lexicon or syntax. Yet, we chose a complex area, tense and aspect, that allows for manifold investigations.

Before we get into more details concerning the methodology and the background of the current study, we need to clarify a number of terminological issues. The first domain that will be addressed is the labeling of the individual languages of a speaker.