Cross-linguistic influence in simultaneous and
sequential bilingual adults.
Judith Selder
S3708624
MA in Applied Linguistics
Faculty of Liberal Arts
University of Groningen
Supervisors:
Dr. Hanneke Loerts
Dr. Marije Michel
Date of submission: 06-07-2019
Word count: 14.321
Declaration of Authenticity
MA Applied Linguistics - 2016/2017MA-thesis Student name: Judith Selder
Student number: s3708624
PLAGIARISM is the presentation by a student of an assignment or piece of work which has in fact been copied in whole, in part, or in paraphrase from another student's work, or from any other source (e.g. published books or periodicals or material from Internet sites), without due acknowledgement in the text.
TEAMWORK: Students are encouraged to work with each other to develop their generic skills and increase their knowledge and understanding of the curriculum. Such teamwork includes general discussion and sharing of ideas on the curriculum. All written work must however (without specific authorization to the contrary) be done by individual students. Students are neither permitted to copy any part of another student’s work nor permitted to allow their own work to be copied by other students.
DECLARATION
I declare that all work submitted for assessment of this MA-thesis is my own work and does not involve plagiarism or teamwork other than that authorised in the general terms above or that authorised and documented for any particular piece of work.
Signed
Date 06-07-2019
Table of contents
Declaration of Authenticity ... 1
Abstract ... 4
1. Introduction ... 5
2. Background ... 8
2.1.1 L2 Acquisition/development and age effects ... 8
2.1.2 Cross-linguistic influence (CLI) ... 10
2.1.3 Difference cross-linguistic influence and transfer ... 14
2.1.4 Mechanisms underlying CLI ... 15
2.1.5 Previous research ... 18
CLI in simultaneous bilingual children ... 18
CLI in adult L2 learners ... 20
CLI in simultaneous and sequential bilingual adults ... 22
2.1.6 Present experiment ... 23
2.1.7 Long passives in Dutch, English, and German. ... 24
2.2 Statement of purpose/ Research questions and hypotheses ... 27
3. Method ... 31
3.1 Participants ... 31
3.1.1 Background measure: LexTALE ... 32
3.2 Materials and design ... 33
3.2.1 Self-paced listening task ... 33
3.2.2 Digitspan task. ... 37
3.2.3 LexTALE ... 38
3.2.4 Language history questionnaire ... 38
3.3 Procedure ... 39
3.4 Analysis ... 40
4. Results ... 42
4.1 Data preparation ... 42
4.2 Linear Mixed Effects Model ... 42
4.2.1 Segment 3 ... 43
4.2.2 Segment 4 ... 44
4.2.3 Segment 5 ... 45
4.2.4 Effect of age of acquisition on RTs. ... 46
4.2.5 Effect of language dominance on RTs. ... 47
5. Discussion ... 48
6. Conclusion ... 55
7.1 Appendix A, List of sentences used in self-paced listening task. ... 58 7.2 Appendix B, Language background questionnaire ... 59 8. References ... 66
Abstract
Cross-linguistic influence (CLI) is a phenomenon that can occur when a person speaks more than one language. Multiple factors have been identified to influence CLI, such as structural overlap and language dominance. Previous research has revealed CLI to be present in simultaneous bilingual children’s speech production and during sentence processing (Serratrice, 2013; van Dijk, Unsworth,& Dijkstra, 2019). What has not yet been investigated, however, is whether CLI is present and, if so, what CLI looks like and which factors influence it in simultaneous and sequential bilingual adults during online sentence processing. Investigating this could reveal more about the potential effect of age of acquisition and language dominance on CLI. The present study aimed at investigating these effects by examining the processing of long passives in Dutch by simultaneous and sequential Dutch-English and Dutch-German bilingual adults. Analyses using linear mixed effects models revealed no CLI to be present and no significant difference between the groups (here: simultaneous versus sequential bilinguals). Additionally, no effect of language dominance was found on CLI. Combined with previous research, these results could suggest CLI to be a developmental phenomenon that is only present during language acquisition. Future research on CLI should compared simultaneous bilingual children with simultaneous bilingual adults and adult second language learners to further investigate this issue.
Keywords: Cross-linguistic influence, simultaneous bilingualism, sequential bilingualism,
1. Introduction
Nowadays, most of the world’s population speaks more than one language (Tucker, 1999). School education plays a large role in this as many secondary schools offer foreign and second language courses (Dutcher, 1994; World Bank, 1995). Students at schools in the Netherlands, for example, are obliged to study foreign languages (i.e. English, French, and German) for the first three years of their secondary school education (Kwakernaak, 2006). Many people, however, grow up learning more than one language at home or elsewhere. For instance, English is present quite prominently in Dutch society. Television series and
commercials are often in English and contribute to a growing bilingual society (Booij, 2001). How languages are stored in the mind remains somewhat of a mystery. Different models have been designed that represent language comprehension or speech production (Levelt, 1989; Zwaan & Radvansky, 1998). Most of these models, however, only focus on monolingual language use. With an increasingly bilingual world, the need to understand how the human mind deals with storing and separating several languages has grown. Recently, the field has gained more interest in how people who speak more than one language (bilinguals) process language and how their languages are stored in the brain. The main focus of these studies has been on the development of languages when children acquire two languages from birth. Two related issues are focused on in particular: language differentiation and cross-linguistic influence (Serratrice, 2013). The first refers to children’s ability to differentiate between their language and treat them as separate and independent systems (Serratrice, 2013; Meisel, 2001;2007). The second refers to the extent in which the two languages of a bilingual influence each other.
It used to be the case that researchers believed that bilinguals acquired their languages relatively separate and that they developed two independent systems (Genesee, 1989; Meisel, 1989; DeHouwer, 1990). However, a shift towards the idea of one shared language system
can be observed at the end of the ‘90s and this gained credibility with the publication of Hulk & Müller’s cross-linguistic influence hypothesis (Hulk & Müller, 2000;Müller & Hulk, 2001). This hypothesis suggested that the two languages of a bilingual can influence each other, especially the interface between syntax and pragmatics is het most vulnerable. An example of cross-linguistic influence comes from a study by Nicoladis (2006) who investigated adjective noun placement in English-French bilingual children. A bilingual child might use a French structure in her English, such as apple green instead of green apple, caused by the influence coming from French.
Since then, multiple researchers have investigated what cross-linguistic influence looks like in bilingual children and what the mechanisms underlying cross-linguistic influence are (e.g. Genesee, Nicoladis & Paradis, 1995; Serratrice 2013; Kidd, Donnelly,& Christiansen, 2018). Cross-linguistic influence has been attested in child’s speech production and during sentence processing (Serratrice, 2013; van Dijk, Unsworth, & Dijkstra, 2019). This phenomenon has also been investigated with respect to the production of speech by simultaneous bilingual, i.e. adult bilinguals who acquired both their language from birth, but no cross-linguistic influence was found there (Kupisch, 2014). What has not yet been investigated, however, is whether cross-linguistic influence occurs in bilingual adults during language comprehension, and, if so, whether cross-linguistic influence differs between simultaneous and sequential (i.e. adult bilinguals who acquired their second language between the ages of 4 and 8) bilingual adults during sentence processing. Investigating this could reveal more about how languages are stored and processed in the bilingual mind and about the development of cross-linguistic influence (CLI). In particular, it could reveal potential effects of age of acquisition (AoA) on CLI by comparing the presence and manifestation of CLI in simultaneous versus sequential bilingual adults.
The present research will investigate cross-linguistic influence in bilinguals who acquired both their languages from birth (simultaneous bilinguals) and bilinguals who learned their second language between the ages of 4 and 8, after they already had their first language system largely in place (sequential bilinguals) once they reached adulthood. This will reveal more about what cross-linguistic influence looks like in bilingual adults and whether the age of acquisition has an effect on cross-linguistic influence. In addition, the present research is also interested in the effects of language dominance on CLI, since it is expected that the simultaneous and sequential bilinguals will shows a difference in their language dominance. This is investigated using a self-paced listening technique and measuring reaction times while processing sentences.
Studying CLI in adults is important because this could reveal more about the way that adults process and store languages in their mind. Investigating the effect of age of acquisition on CLI can contribute to the debate about the effects of age of acquisition on language acquisition and provide a new view on language processing. This way, it can become clear whether age of acquisition has an influence on the more subconscious processes, such as language processing. Additionally, the present research can contribute to the growing evidence of language dominance on CLI.
Following the introduction, an extensive literature review is provided describing the mechanisms underlying cross-linguistic influence as well as findings from previous research on this topic. Next, the method used to investigate cross-linguistic influence is explained containing information about the participants, design of the study, testing procedure, and a description of how the data was analysed. Subsequently, the results are provided and these are discussed and interpreted further in the discussion section. Finally, the conclusion section provides a summary of the main findings and some suggestions for further research.
2. Background
2.1.1 L2 Acquisition/development and age effects
A topic that has been widely discussed in language acquisition research is that of age effects on second language (L2) acquisition and development. Johnson and Newport report on critical period effects in L2 acquisition in one of the most often cited articles in the field (Johnson & Newport, 1989). Initially, it was believed that the critical period hypothesis only had consequences for first language acquisition. Lenneberg (1967) hypothesized that the period in which language could be acquired extended from birth until puberty. Essentially, a late start in language acquisition results in lower levels of ultimate attainment and a difference in language use will remain visible between early and late starters.
Johnson and Newport (1989) then tested whether the critical period hypothesis also extended to second language acquisition. In order to test this, they conducted a study including 46 native Korean and Chinese speakers who moved to the United States and had an age of arrival between 3 and 39 years old. All subjects were tested on a variety of English grammar structures using a grammaticality judgment task. They found that an early start in second language learning lead to higher levels of proficiency compared to late second language learners. Johnson and Newport concluded from these results that a critical period for language acquisition also extends to second language acquisition.
Numerous other studies have been conducted on this topic, revealing similar results (e.g. Coppertiers, 1987; Sorace, 1993). However, the critical period in second language acquisition has also received criticism. For instance, White and Genesee (1996) found that late starters could not be distinguished from native speakers in a grammaticality judgement task. Similarly, Bongaerts et al. (1999) investigated the effect of age of acquisition (AoA) on L2 pronunciation. They demonstrated that the pronunciation of some of the late starters could not be distinguished from native speakers revealing that it is possible for some late learners to achieve native-like levels.
Although there seems to be some controversy on the topic of age effects in second language learning, and especially on whether late learners can reach native-like levels, overall, there seems to be a consensus that there is some effect of age on the development and acquisition of the L2. The question remains, however, what the cut-off point for the critical period is. Most researchers suggest towards puberty for this cut-off point (Lenneberg, 1967; Kroll & Groot, 2005; Birdsong, 2009).
Another issue is the boundary as to when a speaker can be considered a simultaneous bilingual. There is of course a difference between those who learned both languages from birth and those who learned the second language later, but the exact age at which acquisition is considered simultaneous or sequential remains difficult to determine. This is especially the case for child bilingual speakers. The boundary between simultaneous bilingual and sequential bilingual remains rather arbitrary since very few empirical studies address this issue (McLaughlin, 1978). The cut-off point suggested by McLaughlin (1987) is around 4 years since most of the child’s first language is relatively well established at this age. A child acquiring two language from birth is thus considered a simultaneous bilingual, whereas a child acquiring their second language starting after the age of 4 is considered a sequential bilingual. The cut-off point of 4 years to distinguish between simultaneous and sequential bilinguals, which is accepted by many scholars in the field, is also accepted and used in the present study.
Since the present research is only interested in examining the differences between simultaneous and sequential bilinguals it is necessary to define the concept of sequential bilingual further. When can someone be considered a sequential bilingual and when are they considered an adult or late second language learner? As previously mentioned, there is a lot of debate about the age at which the decline in the ability to reach native-like proficiency in a second language starts. For the purpose of the present research the cut-off point of eight years
will be adopted, following Schwartz (2004). The reason for this is that, as Schwartz (2004) notes, in both DeKeyser (2000) and Johnson and Newport (1989; 1991) children who started to acquire a second language before the age of eight were tested on a variety of morphosyntactic phenomena and scored the same as a native speaker, whereas the children that started learning the L2 after the age of eight did not score in native-like ranges. There are thus two cut-off points when it comes to simultaneous, sequential, and late acquisition. The 4 year old cut-off point between simultaneous and sequential bilingualism, and the cut-off point of 8 years between sequential bilingualism and adult or late bilingualism. The definitions used in the present research are thus as follows:
(1) Definition of simultaneous bilingual
A native speaker of two language with initial exposure between 0 and 4 years.
(2) Definition of sequential bilingual
“A non-native acquirer whose initial exposure to the target language is between the ages of four and eight years” (Unsworth, 2005, p7).
(3) Definition of L2 adult
“A non-native acquirer whose initial exposure to the target language is at an age of eight years or older” (Unsworth, 2005, p7).
Cross-linguistic influence (CLI) is a phenomenon that can occur when a person speaks more than one language, such as simultaneous and sequential bilinguals. A definition given by Serratrice is as follows: “Cross-linguistic influence refers to instances in which there is evidence for the effect of one language on the other” (Serratrice, 2013, p4). One language thus influences the other language in some way when CLI occurs. An example is when a bilingual child uses a structure from French in her English, such as apple green. In French, the adjective is allowed in both the pre- and post-nominal position, whereas English only allows for pre-nominal position (Nicoladis, 2006) . Structures such as apple green instead of green
apple might thus be used by French-English bilinguals caused by CLI.
Previous research has identified different conditions that might influence the direction and strength of CLI. These are structural overlap between languages, language dominance,
linguistic complexity, and linguistic domain (Serratrice, 2013).
Previous studies have shown that CLI mainly appears when there is structural overlap between the two languages (e.g., Döpke, 1998; Hulk & Müller, 2000; and Müller & Hulk, 2001). Structural overlap as cause of CLI can be seen as a language internal factor and what structural overlap entails can be demonstrated using Table 1 and 2.
Table 1. Structural overlap
Language a Language b
X X
Y
Table 1 above shows a comparison between two languages; Language a and language b. These two languages have certain properties for a particular linguistic structure. Language a has options X and Y and language b only has option X. Table 2 below shows that when a speaker of language b uses language a they will be more inclined to use option X because this option also occurs in language b.
Table 2. Structural overlap
Language a Language b
X X
Y
Bilinguals might therefore use the structure that is shared between the two languages more often in language a compared to monolinguals. Evidence supporting this claim has been found for multiple language combinations as well as for different structures (Foroodi-Nejad & Paradis, 2009; Hacohen & Schaeffer ,2007; Haznedar, 2007; Nicoladis 2002; 2003).
For example, Nicoladis (2002) examined cross-linguistic influence in compound nouns in French-English bilingual children. The study included 25 monolingual English children, 25 monolingual French children, and 25 French-English bilingual children all between the ages of 3 and 4 years old. Participants were asked to create novel compound nouns. Additionally, they were also tested on their understanding of novel compound nouns (noun consisting of two words). Compound nouns in French can be either left-headed or right-headed, whereas compound nouns in English are always right-headed. Thus, there is structural overlap between these languages. Results suggest that the bilingual children reversed compounds in English more often compared to the monolingual English children. Also, the bilingual children opted for the right-headed option more often than the monolingual French children. No differences between the groups were found for comprehension.
Another condition that can affect the strength and direction of CLI is a language external one, namely, language dominance. It has been observed in multiple studies that CLI is more likely to go from a speakers dominant language into their non-dominant language (e.g., Argyri & Sorace, 2007; Nicoladis, 2006; Pérez-Leroux, Cuza & Thomas, 2011; Yip & Matthews, 2000). Language dominance can be defined as a greater proficiency in one of the two language of a bilingual (Nicoladis, 2002). For instance, Argyri and Sorace (2007)
investigated whether language dominance can determine the strength and direction of CLI in older bilingual children. Experimental data was collected from 32 English-Greek simultaneous bilingual children between the ages of 8 and 9. Sixteen of them were English-dominant and living in the UK and 16 of them were Greek-English-dominant living in Greece. Two monolingual control groups were also included. There was structural overlap between these languages in the sense that Greek allows for pre- and post verbal subject positions, whereas English only allows for a pre-verbal subject position. Participants were administered a grammaticality judgement task.
Results suggest cross-linguistic influence to take place, but only in the English dominant bilinguals. The English dominant bilinguals selected the pre-verbal subject position more often in Greek compared to the Greek monolinguals. This suggests influence from the dominant language English on Greek. The fact that no CLI was found for the Greek dominant bilinguals can be explained by looking at structural overlap and linguistic complexity. CLI tends to go from the language with the less complex structure into the language with the more complex structure, in this case from English into Greek.
The above conditions will be focused on in the present research, because these are more well established. However, to get a good understanding of CLI the conditions of linguistic complexity and domain will be elaborated on briefly as well. The condition of linguistic complexity entails that bilinguals tend to transfer from the language with the less complex structure into the language with the more complex structure for reasons of economy. Multiple studies have found evidence that suggests this in terms of derivational complexity (Gavarró, 2003; Strik & Pérez-Leroux, 2011), presence of ambiguity, or language specific versus language universal strategies (Serratrice et al., 2009).
Finally, it is claimed that the interface of syntax with pragmatics or discourse is especially vulnerable to CLI. It is more likely for CLI to occur when there is an interface with
syntax (Hulk & Müller, 2000; Müller & Hulk, 2001; Serratrice, Sorace & Paoli, 2004; Sorace & Serratrice, 2009).
However, cross-linguistic influence does not always occur as expected. The conditions influencing CLI can interact. When one of the conditions mentioned above is met, this does not automatically mean that CLI will occur and vice versa. This is especially the case when the conditions make predictions in different directions. The strength and direction of CLI is thus not always as straightforward. For example, Argyri and Sorace (2007) investigated the role of language dominance, structural overlap, and linguistic domain in English and Italian structures in English-Italian bilingual children. Various structures were included that either did or did not meet the requirement of structural overlap. An equal amount of English dominant and Italian dominant participants were included in the study. Argyri and Sorace found evidence for CLI in the direction that the surface overlap would predict, namely from English into Italian, but only in the English-dominant participants. Thus, in this study the likelihood of CLI to occur was influenced by structural overlap, but it depended on language dominance whether it actually occurred.
2.1.3 Difference cross-linguistic influence and transfer
Cross-linguistic influence can be either qualitative or quantitative in nature. It can be defined as qualitative when bilinguals find certain structures acceptable and possibly also use these structures in the L2 while they are not produced or accepted by their monolingual peers (Strik & Pérez-Leroux, 2011; Yip & Matthews, 2000). This is referred to as transfer by Paradis and Genesee (1996), which mainly entails negative transfer. On the other hand, quantitative CLI can be seen as a difference in the quantity with which certain structures are used by bilinguals when compared to monolinguals. Bilinguals might go for certain overlapping structures more often than monolinguals (Yip & Matthews, 2000; e.g., Müller &
Hulk, 2001; Nicoladis, 2006; Unsworth, 2012). This can also be seen in the study mentioned earlier by Argyri and Sorace (2007).
The literature concerning adult second language learners mainly uses the term transfer when referring to influence of the L1 on the L2 (VanPatten, 2015; Odlin, 2005; Siegel, 2003). The influence referred to is thus usually one way (L1 to L2) and not cross-language (both ways). It is thus more the case that there is interference from the L1 when using the L2 and not so much a subtle influence. Transfer has a more negative connotation to it (Odlin, 1989). Transfer is often also only in one direction. For instance, in adult second language learners there is transfer from the L1 into the L2 (VanPatten, 2015; Odlin, 2005; Siegel, 2003). The influence of one language on the other only goes in one direction. Cross-linguistic influence, on the other hand, does not necessarily go from one particular language into the other
language, but it can go in both directions.
The present research is concerned with evidence of influence of one language on the other caused by structural overlap. Since it will not consider the possibility of bilinguals producing structures that are not accepted in a certain language and therefore will only look at quantitative CLI rather than qualitative CLI, or negative transfer, influence of one language on the other is referred to as cross-linguistic influence (CLI) throughout this thesis.
2.1.4 Mechanisms underlying CLI
As previously mentioned, predicting the strength, direction, and presence of CLI is not always as straightforward based on the conditions that influence it which might make prediction in different directions. Therefore, the field of child bilingualism faced the challenge of coming up with an underlying mechanism for CLI that could account for the many different findings found in research on CLI in child bilinguals. One explanation that used to be assumed but is now typically rejected is that of language competence. It was previously thought that bilingual children had qualitatively different knowledge than
monolingual children. However, this idea is now rejected as bilingual acquisition is considered to be qualitatively similar to monolingual acquisition (Paradis & Genesse,
1996; Serratrice, 2013; van Dijk, Unsworth & Dijkstra, 2019).
An alternative mechanism that tries to explain CLI in bilingual children and adults is language co-activation (e.g., Döpke, 2000; Hulk, 2000; Nicoladis, 2006; Nicoladis & Gavrila, 2015; Serratrice, 2013; Shook & Marian, 2012). This mechanism entails that it is believed that both languages of a bilingual are always active during sentence processing. When a bilingual is thus processing a sentence in language a, structures in language b also get activated. As a result, when using language a, structures from language b can sometimes be selected. However, this does not entail that this happens all the time. This depends on certain factors, such as the factors previously discussed. The fact that the selection of the non-target language does not always happen, accounts for the variability found in child bilingual research on CLI. Additionally, the theory of co-activation can connect research of bilingualism with second language acquisition since this can account for a general underlying mechanism and model on
how more than one language is stored in the mind.
The occurrence of cross-linguistic influence in this way is explained by Nicoladis’ speech production model (Nicoladis, 2006; Nicoladis, Rose & Foursha-Stevenson, 2010). The study mentioned before by Nicoladis (2006) investigating adjective noun placement in French-English bilingual children is an example. Nicoladis proposes that interference in bilinguals takes place at the level of the lemma. This is where syntactic information is also encoded and the correct language is chosen in case of the bilingual. Co-activation of both languages thus creates a situation where there is competition and potentially the non-target form is selected. She does stress, however, that her model is specific to speech production, but it does provide an example of language co-activation.
Another model that accounts for CLI in both bilingual children and adult second language learners is the shared syntax model by Hartsuiker and Bernolet (2015). It assumes that overlapping structures between languages are shared (Hartsuiker & Bernolet, 2017; Hartsuiker, Pickering & Veltkamp, 2004). As a result, activation of a certain structure in one language leads to additional activation of the shared syntax in the other language and this makes selection of that structure more likely. This can also been seen in Figure 1. The left model represents separate syntax and the right model represents shared syntax. As can be seen in the right model the activation of the genitive ‘s in the L2 English also connects this to Dutch since this structure is shared between these languages.
Figure 1. Hypothetical models for the separate and shared syntax model taken from Bernolet, Hartsuiker, and Pickering (2013).
The following example of the genitive ‘s helps to illustrate this.
(1) Dutch: Oma’s emmer
English: Grandmother’s bucket
As can be seen in (1), the genitive in both Dutch and English can be formed with ‘s.
Because this structure is shared between these two languages, it could be the case that the non-target form in Dutch is selected, but this is not always the case. Moreover, the activation/use of a structure in language a makes the selection of that same structure in language b more likely. This can be seen as a priming effect. Serratrice even suggested that priming could be the main mechanism behind CLI (Serratrice, 2013; 2016).
There are a number of studies that have found CLI in language production and offline sentence processing in young children, but that it disappeared with age when comparing young- to slightly older children (e.g., Argyri & Sorace, 2009; Serratrice et al., 2009; 2012). If language co-activation is a cause of CLI then it can become less visible offline when a structure is acquired. However, if this is truly the mechanism behind CLI then it should still be visible online and persist with age. This has been examined by van Dijk, Unsworth, and Dijkstra (2019) who found evidence of CLI in older simultaneous bilingual children during online sentence processing. This suggests that language co-activation is a mechanism underlying CLI and that it persists with age. However, this has only been examined with bilingual children. Although these children were 8 to 10 years old, their language systems were not completely set in place yet. It has not yet been investigated whether CLI also persists into adulthood.
2.1.5 Previous research
CLI in simultaneous bilingual children
Multiple researchers have investigated CLI in simultaneous bilingual children (e.g. Genesee, Nicoladis & Paradis, 1995; Serratrice 2013; Kidd, Donnelly,& Christiansen 2018). When the term bilingual children is mentioned in the following section, this is referring to simultaneous bilingual children since no research has been conducted on sequential bilingual children. Bilingual children’s languages are considered to develop rather independently (e.g.,Genesee, Nicoladis & Paradis, 1995) and although CLI has been attested in children’s speech production and comprehension (e.g., Serratrice, 2013), language interaction is considered to be quite restricted. What young bilingual children do show is mixing. They tend to be able to differentiate between their two languages and use the appropriate language with the correct interlocutor. However, they often show mixing from their dominant language into their non-dominant language (Genesee, Paradis, & Nicoladis, 1995). When mixing is considered as a form of CLI this is evidence for the claim that CLI mainly goes from the
dominant into the non-dominant language. Mixing occurs when a bilingual child uses elements from both languages in one utterance. This can be seen as a form of CLI, because it shows that the two languages are active and connected.
Evidence for CLI has been found in children’s speech production. For example, Müller and Hulk (2001) investigated object omissions of Romance languages in spontaneous interaction in one Dutch-French, one German-French, and one German-Italian bilingual child and French, Dutch, German, and Italian monolingual children. They found that the bilingual children differed in the extent with which they used object drop in comparison to monolingual children. The Romance languages, in this case French and Italian, can in some cases use pro-drop. This means that the pronoun may be omitted in some cases, especially when they are grammatically or pragmatically inferable. This also means that these languages often have rich inflectional systems. Germanic languages, in this case Dutch and German, do not have this option. In this case, the Germanic language influences the Romance language since the bilinguals used significantly less pro-drop in their Romance languages compared to monolinguals. Thus, it can be said that there is a case of CLI.
Moreover, CLI has also been found in online sentence processing in bilingual children. Van Dijk, Unsworth, and Dijkstra (2019) conducted an online sentence processing experiment with English-Dutch and German-Dutch simultaneous bilingual children. Using a self-paced listening task they investigated the processing of passive constructions in Dutch, German, and English with structural overlap between these languages. Self-paced listening is a method that is not used very often in research. In principle, it is comparable to self-paced reading, which is a more widely used technique. Self-paced listening was first used by Ferreira et al (1996) and Van Dijk, Unsworth, and Dijkstra decided to opt for this method because it is suitable for children with little or no reading skills (van Dijk, Unsworth & Dijkstra, 2019). Participants listen to sentences on a segment-by-segment basis and move to the next segment by pressing a
button. Their reaction times are measured. The results suggest that the bilingual children processed the overlapping structures faster compared to a monolingual control group. The overlapping structure thus facilitated the processing of these structures and this can be seen as a form of CLI.
CLI might look different for children and adults since it is thought that the two languages of a bilingual child develop relatively separately (Genesee, Nicoladis, &Paradis, 1995) whereas adult L2 learners are believed to learn their L2 through their L1. This can thus be seen as a different kind of CLI. This type of CLI is usually defined as transfer. L2 learners transfer from their L1 into their L2. What CLI found in adult L2 learners looks like will be elaborated on in more detail in the next section.
CLI in adult L2 learners
CLI has also been found in adult L2 learners. This, however, is often referred to as transfer (VanPatten, 2015; Odlin, 2005; Siegel, 2003). Where the languages of a bilingual child are considered to develop relatively separate, an L2 adults’ first language (L1) is considered to play an important role during L2 acquisition and their languages have been shown to interact (e.g., Jarvis & Pavlenko, 2008; Hartsuiker & Pickering, 2008). It can thus be said that CLI occurs in adults who are acquiring a second or foreign language as well.
Transfer in L2 (adult second language) learners is found to usually go from the L1 to the L2 (VanPatten, 2015; Odlin, 2005; Siegel, 2003). However, a study by Hohenstein, Eisenberg, and Naigles (2006) suggests that transfer in L2 learners can also go both ways. They investigated cross-linguistic influence in Spanish-English bilingual adults. Their study included two groups of participants; early and late learners of English. The early learners had an AoA (age of acquisition, this is the age of first exposure to a second language) between 4 and 7 years old and the late learners had an AoA after puberty (15 years old). Results suggest that both groups displayed L1-to-L2 grammatical transfer, but both groups also showed
influence of the L2 on lexical decisions in the L1. There was, however, an effect of AoA on both these types of transfer. The grammatical transfer from L1 to L2 was less for the early learners than for the late learners. On the other hand, transfer from the L2 to L1 with regard to lexical choice was higher for the early learners compared to the late learners. This suggests that there was an effect of AoA on transfer. What these results also suggest is that cross-linguistic influence also occurs in L2 learners and that this group does not exclusively show transfer from the L1 to the L2, but that there is cross-linguistic influence and both languages
influence each other.
Hoehenstein, Eisenberg, and Naigles (2006) results, however, do not take into account that the early learners might have had a higher proficiency than the late learners because they had had more years of learning and that this discrepancy in proficiency could have influenced the difference in transfer between these two groups in addition to AoA. Moreover, the participants in their study were all tested at the age between 20 and 25. Meaning, the early learners had had quite a few years more exposure to the L2 compared to the late learners. Therefore, their proficiency is very likely to be higher and this might have been another factor influencing the results.
Another study that confirms that level of proficiency can influence transfer is a study by Sparks, Patton, Ganschow, and Humbach (2009). They investigated transfer from the L1 to the L2 in three groups of learners with either low, intermediate, or high proficiency in their L2. Results suggest that the learners in the low and intermediate groups showed significantly more transfer from the L1 to the L2 than the high proficiency learners confirming that transfer can decrease as proficiency increases.
CLI in simultaneous and sequential bilingual adults
Little research has been conducted on CLI in simultaneous bilinguals and no research can be found on sequential bilinguals once they achieved adulthood. Therefore, it is unclear what CLI looks like in these two groups of bilinguals. A researcher that has investigated language phenomena in simultaneous bilingual adults is Tanja Kupisch. She, for instance, investigated the effects of language context, language attrition, and the vulnerability of individual domains on French in German-French bilingual adults in one of her studies (Kupisch, Lein, Barton, Schröder, Stangen, & Stoehr, 2014). One group of participants grew up in France and moved to Germany during adulthood and the other group grew up in Germany and stayed there. Results indicate that language attrition in adult bilinguals is most prominent in pronunciation and that the environment where the language is learned plays a large role in the development of that language. The group that grew up in France showed higher levels of proficiency, but, at the same time, also more attrition.
Another research by Kupisch (2014) focused on adjective placement in simultaneous German-Italian bilinguals. Adjectives in Italian show variation in their placement with respect to the noun. It looks similar to adjective placement in German, but it is syntactically more complex. The overuse of pre-nominal adjectives was already found in German-Italian bilingual children. Kupisch (2014) thus addressed in her study the question whether adult simultaneous German-Italian learners also showed this overuse of the pronominal position for adjectives in Italian.
Findings revealed that in natural speech production the adults did not show CLI. This suggests that the use of non-target like structures by bilinguals disappears over time.
However, the findings only suggest that CLI disappears over time in production. It does not reveal anything about CLI in language processing and whether this disappears with
simultaneous bilingual adults. This is less of an issue when considering simultaneous
bilingual adults compared to late second language learners. However, simultaneous bilinguals can be more dominant in one of their languages and Kupisch does not take this into
consideration.
A study by Hernandez, Bates, and Avila (1994) did investigate online sentence interpretation amongst one hundred subjects divided over three groups. The first group consisted of 45 Spanish-English bilinguals with a mean age of 20.57. The second group consisted of 25 monolingual speakers of English with a mean age of 19.8 and the third group were 30 monolingual speakers of Spanish with a mean age of 20.86. Participants were administered a sentence interpretation task in which they had to indicate who the agent in a sentence was. Strategies for interpreting such sentences differ for English and Spanish. They found that the simultaneous English-Spanish bilingual adults used sentence interpretation strategies from Spanish when interpreting English sentences. They did so more often than their monolingual peers. This shows the influence of English on Spanish when interpreting sentences and this is thus evidence that implies that CLI still exists in bilingual adults when processing sentences.
However, there is very little evidence to support this claim and therefore more
research into CLI during sentence processing in simultaneous and sequential bilingual adults is necessary. Investigating this can reveal more about whether CLI occurs in these groups and about the effects of age of acquisition and language dominance on CLI.
2.1.6 Present experiment
The effects of simultaneous or sequential acquisition of language on CLI is something that has not yet been investigated. It is unclear how CLI develops over time and whether a difference in CLI can be seen between these two groups, caused by differences in age of acquisition. The development of CLI overtime in itself is a phenomenon that has not yet been
investigated in previous research. Most previous research on CLI has focussed on simultaneous bilingual children.
The present research will thus focus on both simultaneous and sequential bilingual adults and whether the difference in age of acquisition influences the presence of CLI found, if CLI is found in these two groups in the first place, during online sentence processing. Since the effects of a later age of acquisition remain visible in language acquisition (Coppertiers, 1987; Johnson & Newport, 1989; Sorace, 1993) it is expected for these effects to also appear on CLI. The sequential bilinguals will thus probably show more CLI compared to the
simultaneous bilinguals. Additionally, the present research will also investigate the effects of language dominance on CLI further.
CLI might also be a developmental phenomenon. Meaning, that it could be present in childhood but disappears over time (Kupisch, 2014). However, it might also be the case that it does not disappear over time, but that it is mainly present during language acquisition. Since children are still acquiring their languages and adults already have their language system in place this difference in CLI might be caused by the process of acquiring language. The present research thus also indirectly investigates this. When no CLI is found in adult bilinguals who have their language systems in place, this could suggest CLI to be a developmental phenomenon. The effect of age of acquisition will thus be investigated, but also whether it is a developmental phenomenon of language acquisition.
2.1.7 Long passives in Dutch, English, and German.
CLI in simultaneous and sequential bilingual adults will be investigated by looking at structural overlap between Dutch, English, and German and by looking at language dominance (proficiency levels). Structural overlap occurs in long passives in Dutch when compared to English and German. The following structures occur in Dutch:
(2) past participle – verb (PP-V) De beer wordt door de leeuw geduwd. The bear is being by the lion pushed The bear is being pushed by the lion
(3) V-PP De beer wordt geduwd door de leeuw.
The bear is being pushed by the lion
As can be seen, the by-phrase is relatively flexible. It can either precede (PP-V) or follow (V-PP) the past participle. Both options are grammatical, but they differ in the information they emphasise. In (2) the agent of the sentence is emphasised whereas in (3) the emphasis is placed on the action. This locates the passive in Dutch in the syntax-discourse interface,
making it more vulnerable to CLI.
Typical surface word order in Dutch main clauses is SVO (Zwart, 2011) and therefore the V-PP word order is probably less marked than the PP-V word order. The use of both these passive constructions by Dutch monolinguals has been examined by Bernolet, Hartsuiker, and Pickering (2013). They collected frequency data on the use of the different transitives in a picture description experiment. The results for the passive constructions were as follows: 10.6% of transitive sentences were described with passive sentences and of those were 4.9% PP-final passives and 2.4% PP-medial passives. The other 3.3% were short passives in which no agent is expressed (Bernolet, Hartsuiker, Pickering, 2013). Although the final and PP-medial passives are not used the same amount of times, an acceptability judgement task revealed that native speakers of Dutch consider both options equally acceptable (Bernolet, Hartsuiker, Pickering, 2013).
For English and German, however, the word order of long passives is more fixed. The by-phrase in English always follows the past participle (Fox & Grodzinsky, 1998) and in German the past participle always occurs in the final position (Beedham, 1981).
For both languages, it is the case that either the V-PP structure or the PP-V structure is accepted to be grammatical. For English this is the V-PP structure and for German the PP-V structure.
English:
(4) PP-V *The bear is being by the lion pushed. (5) V-PP The bear is being pushed by the lion. German:
(6) PP-V Der Bär wird vom Löwe geschubst.
The bear is being by the lion pushed The bear is being pushed by the lion
(7) V-PP *Der Bär wird geschubst vom Löwe.
The bear is being pushed by the lion
It would thus be expected that Dutch-English simultaneous bilinguals and English sequential bilinguals learning Dutch as a second language will prefer the V-PP option in Dutch, because this structure overlaps between the two languages as can also be seen in Table 3. On the other hand, it is also expected that Dutch-German simultaneous bilinguals and German sequential bilinguals learning Dutch as a second language will prefer the PP-V option in Dutch, because of structural overlap between these languages as can be seen in Table 3. They will use these structures more often than monolingual speakers of Dutch.
Table 3 – language overlap between English and Dutch (V-PP) and German and Dutch
(PP-V)
English Dutch German
PP-V PP-V
V-PP V-PP
However, when considering language dominance as a predictor of CLI it is also expected that the sequential bilinguals will show a stronger preference for the option also
available in their L1 compared to the simultaneous bilinguals, because the sequential bilinguals are expected to be more dominant in their L1. Dominance will be measured by looking at proficiency levels in both languages and subtracting the English or German proficiency score from the Dutch proficiency score. This way it will become clear which language is the more dominant one, because the dominant language will be the language with greater proficiency. This way of looking at language dominance was also administered by van Dijk, Unsworth, and Dijkstra (2019).
2.2 Statement of purpose/ Research questions and hypotheses
Based on this literature review the following gaps in the literature were discovered. CLI in children has been widely examined. Similarly, transfer in second language learners has also been investigated. However, very little is known about CLI in simultaneous and sequential bilingual adults. Previous research only suggests that CLI in speech production can decrease with age, as based on a comparison between simultaneous bilingual children and adults. Yet, what has not yet been investigated is whether CLI in sentence processing also disappears over time or whether it is still present in adulthood. Since CLI is also found in second language learners it could be the case that it is a phenomenon of language acquisition. Meaning that it is mainly present when language acquisition is taking place.
On the other hand, since age of acquisition plays an important role in language learning and processing, it could be the case that AoA influences presence and type of CLI.
Comparing CLI in simultaneous and sequential bilingual adults has the following advantages:
CLI in sentence processing has not yet been investigated in simultaneous and sequential bilingual adults.
CLI resulting from having two languages can be teased apart from CLI resulting from learning a language after having acquired one already.
It will become more clear whether age of acquisition has an effect on CLI.
It will become more clear what the effect of language dominance/level of proficiency is on CLI.
This will be investigated with the following research questions: 1. Does CLI occur in simultaneous and/or sequential bilingual adults?
There have been various studies on this topic that put forward opposing views. For instance, Kupisch (2014) found that simultaneous German-Italian bilingual adults did not show CLI in natural speech production. On the other hand, Hernandez, Bates, and Avila (1994) investigated online sentence interpretation in Spanish-English bilinguals and did find evidence for CLI in adults during sentence processing. Also, van Dijk, Unsworth, and Dijkstra (2019) found evidence of CLI during online sentence processing in older bilingual children (8 to 10 years old). Since the present research is interested in CLI during sentence processing and previous studies that investigated CLI during sentence processing in older bilingual children and adults did find evidence of CLI, it is expected for CLI to occur in simultaneous and sequential bilingual adults during online sentence processing.
Whether CLI occurs in simultaneous and sequential bilinguals can be examined by looking at Dutch and by comparing the bilinguals with monolinguals. The overlapping structures between languages will be processed faster and the unknown structures will be processed slower.
If the answer to the first research question is ‘yes’, then the second question to address is:
2. Does simultaneous compared to successive acquisition affect the presence and/or type of syntactic cross-linguistic influence in bilingual adults?
The hypothesis is that a difference in CLI between the two groups will be visible. Sequential bilingual adults will probably show more cross-linguistic influence in language processing in comparison to simultaneous bilingual adults, because they have a later age of acquisition (Coppertiers, 1987; Johnson & Newport, 1989; Sorace, 1993). The overlapping structures will be processed faster and the unknown structures will be processed slower.
The effect of age of acquisition can be checked by comparing one group of German-Dutch simultaneous bilinguals to one group of German-German-Dutch sequential bilinguals.
3. To what extent does language dominance affect the presence of cross-linguistic influence?
It is expected that language dominance will also influence the amount of CLI. Previous research has revealed that CLI goes from the dominant language into the non-dominant language and that CLI tends to get less with higher proficiency (Argyri & Sorace, 2007; Nicoladis, 2006; Pérez-Leroux, Cuza & Thomas, 2011; Yip & Matthews, 2000. Sparks, Patton, Ganschow, & Humbach, 2009). Therefore, it is expected that subjects that are more dominant in Dutch show less CLI (Hoehenstein, Eisenberg, & Naigles, 2006; Sparks, Patton, Ganschow, & Humbach, 2009).
It is important for the purpose of the present research to tease apart age of acquisition and language dominance. Therefore, to determine the age of acquisition results from a questionnaire will be analysed and the group in which a participant belongs will be determined on the basis of their answers to this questionnaire. The age of acquisition is the age of first exposure to Dutch. On the other hand, a bilingual’s dominant language is not necessary their first language (Kupisch, 2008). Therefore, in the present research it is not assumed that the first language of the sequential bilinguals is also their dominant language.
This is tested by looking at proficiency levels of Dutch and English or German. Since the dominant language is usually the language with the greater proficiency (Nicoladis, 2002), the English or German scores will be subtracted from the Dutch scores. The design of the present study will also take this into account.
3. Method
The present study was designed as a follow up study to a previous study conducted on this topic with bilingual children by van Dijk, Unsworth, and Dijkstra (2019). Their study used a similar method, however, this has been adapted in order to be suitable for research with adults and to see whether the results found in the study with children were also replicable with adults. Moreover, the present study was also combined with another study on cross-linguistic influence operating the same self-paced listening method, but with a different focus and research question unrelated to the current study. The methods and results of this second study will not be discussed as the sentences from this study only served as the fillers of the present study and the other way around. This study will be referred to as study2 throughout the rest of this method section.
3.1 Participants
Three groups of participants were included in this study: simultaneous bilingual adults, sequential bilingual adults, and a control group of Dutch monolinguals. The simultaneous bilingual adults consisted of 13 German-Dutch and 11 English-Dutch bilingual adults between 15 and 30 years old. These participants were simultaneous bilinguals who acquired both their languages before the age of 4. The mean age, standard deviations, mean age of acquisition,
and range are provided in Table 4.
The sequential bilingual group consisted of 16 German-Dutch bilinguals between 15 and 30 years old. These participants were sequential bilingual adults who started acquiring Dutch between the ages of 4 and 8 years old. Their native language was German and they learned Dutch as a second language either at home or in school. Criteria for inclusion for both the simultaneous and the sequential bilinguals were that they currently had a minimum of half a day a week of exposure to both languages. Limitations on past exposure were that participants did not study a language programme and had not lived abroad for more than three
months. This was tested using a questionnaire (see Appendix B). Based on the answers provided by the participants, it was decided whether or not to include the participant. The control group consisted of 14 monolingual speakers of Dutch between 15 and 30 years old. It is relatively hard to find monolingual speakers of Dutch in the Netherlands since everyone has to learn a second language at school and because English is present quite prominently in Dutch society. Therefore, the participants were asked to specify how much contact with and exposure to any second or foreign they had at the time of testing (especially German and English) and how much second language instruction they had previously had. Criteria for inclusion were that they did not have significant exposure to a second language before the age of 10 (group 7 in the Dutch school system), had not lived abroad for more than three months, and that they currently only had passive exposure to an L2 for no more than one day a week. Participants that exceeded this threshold were not included in the experiment. None of the participants suffered from any specific language impairment or hearing problems.
Table 4. Mean age and standard deviations and mean age of acquisition and range for the
monolingual, simultaneous German, simultaneous English, and sequential German groups.
Monolinguals Simultaneous-Germans Simultaneous-English Sequential-Germans Mean age (standard deviations) 23.1 (4.2) 25.3 (8.1) 28.9 (5.6) 26.4 (7.8) Mean age of acquisition (range) >1 (0) 1.6 (0-3) 1.9 (0-3) 6.7 (5 – 8)
3.1.1 Background measure: LexTALE
Table 5 below provides the results of the LexTALE proficiency task. It shows the means and standard deviations of the relative dominance scores (Dutch scores – English or German scores) for the groups of simultaneous German bilinguals, simultaneous English bilinguals, and the sequential German bilinguals.
Table 5. Means and standard deviations of the relative dominance scores for the simultaneous
German, simultaneous English, and sequential German groups.
Simultaneous German Simultaneous English Sequential German Mean dominance score (Standard deviation) 6.9(7.3) -7.5 (8.4) -13.5 (14.3)
As can be seen in Table 5, it seems to be the case that only the simultaneous German group is Dutch dominant. Whereas the simultaneous English group is English dominant and the sequential German is German dominant. The standard deviations also suggest that there is a lot of variation within the groups regarding dominance scores, with the largest variance in the sequential German group (sd = 14.3).
3.2 Materials and design 3.2.1 Self-paced listening task
A self-paced listening experiment was designed to measure participants’ online
sentence comprehension (Ferreira, Henderson, Anes, Weeks & McFarlane, 1996). During this task, participants were asked to listen to sentences segment by segment. By pressing a key on a button box, they moved on to the next segment. The self-paced listening method was opted for, because this study was part of a larger set of studies also including studies with young children. Because young children often have little or no reading skills (Marinis, 2010), self-paced listening is a good alternative to self-self-paced reading.
Other advantages of self-paced listening over self-paced reading are that self-paced listening requires less focus since participants are not dependent on looking at a screen to receive the information that is being presented. A critique that self-paced listening often receives is that it can sound unnatural (Papadopoulou, Tsimpli, Amvrazis, Jegerski, &
VanPatten, 2013). However, when the sentences are recorded as a whole and only segmented afterwards the natural intonation of the sentences is kept intact. Moreover, as mentioned by
Jegerski (2014), self-paced reading has the disadvantage that there might be great individual differences in reading speed. When measuring reaction times this could influence the results (Jegerski, 2014). Since the present study will be considering reaction times, self-paced listening is a suitable method that eliminates individual differences in reading speed.
A total number of 70 sentences were created: 10 practice items, 30 experimental items that were all long passives, and 30 active sentences which were included and used as filler sentences to distract the participant from the purpose of the study. Besides these fillers, the items from study2 also served as filler items and were unrelated in structure to the
experimental items of the present study. Of the experimental items 15 items were PP-V structure and 15 were V-PP structure (also see Table 6). The active filler sentences were created in such a way that they were structurally unrelated to the experimental items. These filler items consisted of the following sentence structures: 10 sentences with dative case marking, 10 sentences with pp-attachment, and 10 sentences with particle verbs. Examples of sentences are provided in Table 6. A list of all sentences used can be found in Appendix A.
All sentences were created following a set of criteria.
All verbs were taken from a previous study that considered the acquisition of passives by children (Armon-Lotem et al, 2016);
All nouns were animals with the determiner ‘de’;
All nouns were very frequent animal names in Dutch, because previous researchers
have shown that frequency of words influences sentences processing in listening (Papadopoulou et al., 2013). Therefore, it was ensured that all words were high frequency words in Dutch. They were all taken from a frequency list created for the Dutch lexicon project (Keuleers, Diependaele & Brysbaert, 2010; CELEX – Baayen et al., 1995);
Sentences were created in such a way that the animals’ actions were plausible considering the type of animal.
The long passives consisted of seven segments (also see Table 6). Segment three and four being the critical region and segment five being the spill-over region. When measuring reaction times, the effect sometimes only shows on the word after the critical region and therefore the spill-over segment is considered as well (Smith & Levy, 2013). All sentences were recorded using a female native speaker of Dutch. Recording was done in a soundproof booth using the programme Audacity (Audacity team, 2019). The sentences were recorded as a whole sentence and segmented afterwards using Audacity (Audacity team, 2019). This way, the natural intonation of the sentences was kept intact and processing demands for the
participants were kept as low as possible. The active filler sentences consisted of 4 to 7 segments.
Besides these sentences, 72 filler sentences from study2 were also added to the self-paced listening task. The structure of these sentences was unrelated to the structure of the experimental items of the present study. A total of180 sentences was used for the self-paced listening task.
Table 6 – Example sentences of long passives and active sentences used in self-paced listening task. Segment Long passive 1 2 3 4 5 6 7 PP-V De zebra The zebra wordt is door de pony by the pony geknuffeld hugged en and de vogel the bird
kijkt naar een stoel
watches a chair V-PP De zebra The zebra wordt is geknuffeld hugged door de pony by the pony en and de vogel the bird
kijkt naar een stoel
watches a chair
Comprehension question: Kijkt de vogel? Watches the bird?
1 2 3 4
Active Dative Het varken
the pig geeft gives het paard the horse een knipoog a wink PP-attachment De hond The dog bezoekt visits de gans the goose in haar nest in her nest
Particle verb De spin
The spider zwaait waves de bever the bever uit goodbye
Comprehension question: Zit de gans in haar nest? Is the goose in her nest?
Comprehension questions were asked following 8 of the long passives and 8 of the active sentences (equal number of questions needing a yes or no-response) to ensure that the participants were paying attention. The comprehension questions did not focus on the critical region itself, but rather the action in general using the same verbs and nouns to keep
processing demands for the participant as low as possible. This way, it was avoided that the participants paid particular attention to the structures that were being studied. The active sentences consisted of four to seven segments and comprehension questions were also included for these sentences.
Pseudo-randomized lists were created containing the stimuli. A total of four lists were created each containing all items in a randomized order. All lists started with 10 practice items followed by a filler item and ended with an experimental item. There was a maximum of four consecutive items from study2 and a maximum of three consecutive items from the present study. Fillers occurred in a maximum of two consecutive items. The order in which
the sentences were presented was pseudo-randomized in such a way that there were never more than three items following each other with a comprehension question and four items following each other without a comprehension question. The experiment was created in E-Prime, version 2.0 (Schneider, Eschman & Zuccolotto, 2002). All participants listened to all 180 sentences in a randomized order.
3.2.2 Digitspan task.
The results of the Digitspan task were not included in the present research, but this test was included as a background measure for Study2. It is included in the method of the present study, however, because it was included in the procedure of the experiment and because the results of the Digitspan task might be interesting for future research.
Since short-term and working memory abilities might interact with a person’s sentences processing abilities (Lewis, Vasishth & Van Dyke, 2006), a Digitspan task was employed to test short-term and working memory. The Digitspan task used was based on the Automated Working Memory Assessment (AWMA) (Alloway, 2012). During this task, participants had to repeat digits first in forward order (short-term memory test) and then in backward order (working memory test). The sequences of digits were presented in blocks of four sequences. If a participant repeated four sequences correctly, the next block of sequences started automatically. When a participant made an error and thus three or less sequences within the same block were correct, the experiment would move on to the backward condition. When three or less sequences were correct in the backward condition the task
would end.
The task was administered using E-prime (Schneider, Eschman & Zuccolotto, 2002). Scores were calculated separately for both tests, since the forward condition measured short-term memory and the backward condition measured working memory. One point was rewarded for each correct response.
3.2.3 LexTALE
Proficiency levels will be used to asses a participant’s language dominance, since they are likely to be more proficient in their dominant language compared to their non-dominant language.
In order to test participant’s proficiency of Dutch and English or German the LexTALE online proficiency test was used. This is a readily available online vocabulary test created by Lemhöfer and Broersma (2012) which measures vocabulary size, and it thus provides a rough indication of proficiency level. It is a relatively short test which takes approximately 4 minutes for the participants to complete. The advantage of the LexTALE proficiency task is that it is available in English, Dutch, and German. Therefore, it is comparable across languages and suitable for the purpose of the present research.
One dominance score is calculated for the participants by subtracting the English or German score from the Dutch score. A positive score thus means Dutch dominant. The outcome of the LexTALE is the percentage of correct responses.
Although the LexTALE is only a short test the outcomes are comparable to outcomes of the Quick Placement Test (QPT) and the TOEIC test, which are extensive, thorough, tests of English proficiency (Lemhöfer & Broersma, 2012).
3.2.4 Language history questionnaire
For the purpose of the present study it was important to get a full picture of the previous and current language exposure and language use use of the participants. In order to measure this a combination of questions from the language history questionnaire developed by Li, Zhang, Yu, and Zhao (2019) and the BILEC, which is a questionnaire designed by Unsworth for bilingual children (Bilingual Language Exposure Calculator; Unsworth, 2013), were used. This questionnaire can be found in Appendix B.
The questionnaire was filled out by the participants to assess their language background and language use. These existing questionnaires were adapted to fit the relevance of the present
