2 Definitions, Terminology and Concepts on BFLA
2.2 Definition of BFLA
2.2 Factors Determining BFLA: Age, Input, Access to UG
2.3 Concepts and Theories on BFLA
3 Interdependent Development - Cross-Linguistic Influences in BFLA
3.1 Cross-Linguistic Influences: Definition
3.2 Predicting CLI
4 Empirical Studies showing CLI
4.1.1 Word Order in Subordinate Clauses in French/Italian-German Bilinguals
4.1.2 Wh-Interrogatives in a Cantonese-English Bilingual
4.1.3 Wh-interrogatives in English-Japanese Bilinguals
4.2.1 Determiners in German-French and German-Italian Bilinguals
4.2.2 Verb Position in Main Clauses in German-Italian Bilinguals
4.3.1 Subject drop in German-Italian Bilinguals
4.3.2 Object drop in German-Italian and German-French Bilinguals
5 Interpretation of the Results and Theoretical Implications
5.2 Discussion of Results: Cross-linguistic Influences -
Language-internal vs. Language-external Factors
5.3 General Implications on Language Acquisiton
In today’s society more and more children grow up in a multilingual environment and acquire two languages from birth or soon afterwards. This can be regarded as an advantage and as a chance for these children to acquire more than one mother tongue. On the other hand, even today there are concerns about exposing children to two or more languages from birth since it might confuse them linguistically, cognitively and emotionally. Parents fear that the language development of their children might be delayed and that the children might not be able to keep their two languages apart. This may lead to incomplete knowledge of both languages. (Meisel 2004: 91) Increased linguistic and psycholinguistic research on bilingualism during the last 30 years has helped to eliminate some concerns and aided in our understanding of the linguistic behaviour of bilingual children. The development of the two morphosyntactic systems in particular has been of interest for research, and they examined whether the children could separate their two languages in the course of acquisition or if the systems are influencing each other. This paper will seek to show that there is evidence for cross-linguistic influences on the level of competence in BFLA and that a combination of language-internal and language-external factors can account for this development. First, researchers assumed that children cannot keep their languages apart and argued for a gradual separation. This concept is called the Single System Hypothesis. Until the late 1980s and 1990s, research on Bilingual First Language Acquisition (henceforth, BFLA) developed the view that the children acquire two separate linguistic systems from the beginning, and that there is no language influence on the level of competence. This development is called Autonomous Development. Recently, these two concepts are no longer regarded as completely opposing each other. Some researchers assume that there is interaction between the two morphosyntactic systems of bilingual children in the course of language acquisition. These interactions are called cross-linguistic influences, which cause different acquisition patterns from those of monolingual children, called Interdependent Development. This view is controversial, as some argue that these influences are not manifested at the level of competence. Furthermore, the question remains which factors account for this development.
To create a background for the discussion, the term bilingualism will be defined and a brief overview on concepts and theories in BFLA will be given. Many different definitions of bilingualism can be found in the literature. As this study deals with the simultaneous acquisition of two languages from birth, there will be a focus on this domain, which is called BFLA (Meisel 1989, De Houwer 1990). In the following section, the factors influencing BFLA will be examined as they determine the development of the systems. The age at which children are first exposed to the second language influences their pattern of acquisition. Hence, it is necessary to outline the role of age for the outcome of this process, as the simultaneous acquisition from birth can be characterised as multiple first language development. If this is the case, the universals of First Language Acquisition (henceforth, FLA) could apply in a multiple way. These theories need modification to be transferred to BFLA. Theoretical approaches on the access to Universal Grammar (UG) and to the question of whether the human language faculty is capable of multilingual acquisition will be discussed. UG is activated by input of the target language, and in the bilingual situation the input consists of two languages. The question arises how children differentiate between their languages and what the results of the limited input compared to a monolingual child are. The different amount of input may lead to dominance of one language, which can be regarded as a factor determining cross-linguistic influences. Especially, the storage and separation of the two linguistic systems has been an issue of research for the last decades. There will be a short overview of how the issue has been treated in the literature in the past. It was not always assumed that the two linguistic systems develop separately but that they start out as one system which is differentiated later in development. Volterra and Taeschner (1978) suggested a three-stage model to characterise this development of the linguistic systems of a bilingual child. This model has proved highly controversial, both for empirical and theoretical reasons. Particularly the works of Meisel (1989) and Genesee (1989), written as a reaction to the model, have proved it wrong and have shown that there is an early separation of systems.
However, the form which the relationship between the two developing systems takes remains a controversial topic in literature. One position holds the view that the development takes place without cross-linguistic influences. The second position claims that there are quantitative and qualitative differences compared to monolingual FLA which are due to cross-linguistic influences in certain domains. Autonomous and interdependent development will be defined according to Paradis and Genesee (1996). These definitions will be used to test whether there is evidence for Interdependent Development in the morphosyntactic development in BFLA. Different methodologies and explanations have been proposed to predict the domain and the language influenced in BFLA. Some argue in favour of language-internal factors being the reason for this influence, e.g. Döpke (1998) and Müller et al (2002, 2006). They have defined criteria to predict the domain affected by cross-linguistic influences and which language will influence which. Others argue for language-external factors such as dominance or individual learner types. These methodologies will be theoretically explained and described.
In the next chapter, studies will be introduced which provide evidence for the three potential forms of interdependent development in BFLA: transfer, acceleration and delay. Furthermore, the applicability of the methods previously introduced to explain these patterns will be tested.
In conclusion, there will be general evaluation on the methodology of the studies presented in this paper. Furthermore, the data and results of the studies will be compared to present an overall picture of the studies to provide evidence for Interdependent Development. The reasons for this development will be critically examined to suggest an explanation which combines language-internal and language-external factors. The domain affected seems to be determined by language-internal factors, but the question of whether this development can be observed in the individual is determined by individual language-external factors. Finally, some general implications on the process of language acquisition will be made which allow phenomena such as cross-linguistic influences and provide a contribution to theories on monolingual FLA.
2 Definitions, Terminology and Concepts on BFLA
Bilingualism has been of interest for linguistic research for many years. In this chapter, an introductory overview on BFLA will be given to understand the discussion on which this paper is based. First of all, the term "BFLA" will be defined and problems resulting from different definitions will be discussed. Next, the most important factors influencing BFLA will be presented as they determine the linguistic development of a bilingual and deliver the theoretical basis for discussion. Finally, a short overview will be given on the discussion on the linguistic development in BFLA with focus on the issue of separation of the two systems.
2.2 Definition of BFLA
The term bilingualism has many different facets and is used very differently in the literature. This section will introduce and discuss the different distinctions to outline the special features of BFLA. One of the basic distinctions to be made is between individual and societal bilingualism (Baker 2006: 2). This paper deals with individual bilingualism and not with the phenomenon within a certain group or society. But even with this restriction to the individual level, it is not easy to define the term bilingualism or to determine who is bilingual and who is not. Bilingualism involves very complex cognitive and linguistic processes and varies within different individuals. It is not simply about two languages as monolingualism is about one (Baker 2006: 2). Hence, definitions are different and problematic. While some authors put emphasis on the underlying competence, such as Bloomfield (1935: 56), who defined a bilingual as a person who has “native-like control of two languages”, others, such as Weinreich (1953: 5) prefer a wider definition by putting emphasis on use, and call bilingualism the “practice of alternatively using two languages”. Haugen (1953: 7) goes even further in his definition: for him individuals who are fluent in one language but who “can produce complete and meaningful utterances in the other language” are bilinguals. Beatens Beardsmore (1982: 3-4) points out the problems of these definitions as they reach from a very broad to a very narrow spectrum. For Bloomfield (1935) the degree of proficiency has to be very high and furthermore, the problem of defining “native-like” arises. He even admits himself that it is impossible to “define a degree of perfection at which a good foreign speaker becomes a bilingual: the distinction is relative” (Bloomfield 1935: 56). Haugen (1953), on the other hand, includes everyone who is able to say just a word or a phrase in another language. Today, there is a tendency to adopt a broader view and bilinguals are defined as “individuals or groups of people who obtain communicative skills, with various degrees of proficiency, in oral and/or written forms, in order to interact with speakers of one or more languages in a given society” (Butler and Hakuta 2004: 115). This includes many speakers who may also be in a process of acquiring a second language; hence, it is necessary to classify individual bilingual speakers into different categories.
There are certain key variables needed to classify and to define a bilingual person which are according to Wei (2007: 5): “age and manner of acquisition; proficiency level in specific languages; domains of language use; self-identification and attitude.” According to these variables a large number of different categories for individual bilinguals can be defined. This paper focuses on bilingual children who acquire two languages simultaneously from birth or soon afterwards, a process which is called BFLA. The term was introduced by Meisel (1989) and defined more precisely by DeHouwer (1990: 3):
“BFLA refers to these situations in which
a) a child is first exposed to language B no later than a week after he or she was first exposed to language A, and
b) a child’s exposure to language A and B is fairly regular, i.e. the child is addressed in both languages almost every day”
She places emphasis on the time when the child is exposed to the second language, and sets a very narrow time frame when the exposure to the second language has to start. This will be discussed at greater depth in 2.2. In the following, categories relevant for these bilinguals will be introduced which are frequently used in the literature.
The children are exposed to two languages from birth, so this form is sometimes called early or infant bilingualism. They acquire these languages in a naturalistic context which means that they have bilingual or multilingual parents or grow up in a bilingual or multilingual community. This can be contrasted to adults or children who acquire a second language successively through formal instruction (Wei 2007: 5). Hence, BFLA is labelled simultaneous and natural bilingualism.
When measuring the proficiency level, some distinctions have to be taken into consideration. There has to be a distinction between the performance of a speaker and the underlying competence (Baker 2006: 3). The outcome of BFLA should be native-like competence in both languages (Müller et al. 2006: 13). When the relationship between the competence in the languages is equal, which means that they are fluent in both languages, the form of bilingualism is called balanced (also: equilingual, ambilingual). When there is higher proficiency in one language, the child is a dominant bilingual. (Butler and Hakuta 2004: 115) Although children acquiring two languages from birth should be fairly balanced, it frequently occurs that a child is dominant in one language. But it is necessary to state that BFLA does not imply the same frequency of usage but “that both input languages start to be used in regular communication with the child at the same time in development” (De Houwer 2005: 31). Through individual circumstances such as the surrounding language environment, the child can temporarily become more proficient in the other language.
2.2 Factors Determining BFLA: Age, Input, Access to UG
The factor of age is very important in BFLA, as becomes evident when looking at studies and theories from Second Language Acquisition (SLA). (Meisel 2004: 105) Lenneberg (1967) developed a hypothesis which claimed that the language capacity is only accessible during a limited age period. This is called the Critical Period Hypothesis (CPH), which describes three periods of language acquisition. The first period starts with an abrupt onset with a peak before 2;0 (2 years and 0 months) and gradually subsides until the age of 5;0. Afterwards there will be qualitative differences in the patterns of acquisition. This results in three types of bilingual acquisition (Meisel 2004: 105):
1. simultaneous acquisition of bilingualism (2L1) - both languages start being acquired during the first three or four years.
2. child second language acquisition (L2) – the second language starts to be acquired between 5 and 10 (puberty).
3. adult second language acquisition (L2) – the acquisition starts after the age of 10 (puberty).
This shows that according to the CPH, it is still regarded as simultaneous acquisition when the child is exposed to the second language before the beginning of the subsidance. Whether this is true or whether this leads to competence differences as compared to children who were exposed to the two languages from birth on remains controversial. Meisel (2004: 106) speculates that exposure during the period of optimal age is necessary but not sufficient for native-like competence. The reduction of input of one language in early childhood may lead to deficits. Arguing this way, it is difficult to draw exact boundaries between simultaneous and successive child bilingualism, as it seems to be possible to acquire a language without formal constructions in nursery or kindergarten (Baker 2006: 97). As these theories and assumptions are still controversial amongst linguists, for the purpose of this paper it is sensible to use the definition of DeHouwer (1990) and require that the child be exposed to the second language not later than a week after birth.
In child SLA, the child already has linguistic knowledge in the initial state and this can be activated for the acquisition of the second. Other differences are the order of sequences, as they are not identical in L1 and L2 acquisition. (Meisel 2004: 106-7) Whether children are better learners and whether only they are able to gain native-like competence in a second language, and whether the critical period for language acquisition ends with puberty remains controversial but this will not be discussed here. Nevertheless, from a biological point of view the neurophysiologic characteristics of the human language faculty start to change at the age of one and the brain looses its plasticity at the end of puberty. Furthermore, it was shown that after the age of three the second language is stored differently in the brain (Obler et al. 1982: 458) and that consequently there has to be a different pattern of acquisition (Müller et al. 2006: 14).
Furthermore, it has to be kept in mind that when children start to learn two languages from birth, they do not only learn to speak and to use two languages, but learn that language is a way of expression and communication. They acquire formal aspects of two languages such as sounds, words, meanings, relationships and its grammar at the same time. (Hoffmann 1991: 34) This means that they do not simply acquire native-like linguistic competence in both languages, but also acquire knowledge as to how the languages are used in different social contexts and situations from the beginning onwards (Hoffmann 1991: 35).
The point in time at which the child is exposed to the second language also determines the amount of input. Input is a very important factor in language acquisition in general. Primarily, input is the actual linguistic form the native speaker uses with the learner, consisting of sounds, words, and utterances. Additionally, the manner of presentation of those forms and the metalinguistic information are part of the input (Schachter 1986: 215). Lanza (1988: 70) emphasises that the cultural content also plays a role which is of special significance in bilinguals, as the two different languages are mostly embedded in different cultural contexts. The utterances of the target languages spoken to the child are analysed, and by means of these analyses grammatical rules are established. There are several problems concerning input, not only in BFLA but also in monolingual FLA. The input which the child is exposed to is only a small and incomplete excerpt of the target language. Some structures occur only rarely in daily language usage, nevertheless, children acquire these structures correctly. (White 2003: 4) This problem is called poverty of the stimulus, and leads to the assumption that there is an innate language capacity which helps children to acquire grammatical rules.
Input sometimes allows more than one interpretation, and the child might develop more than one analysis at the same time. Müller et al. (2006: 29) state that there are different hypotheses on strategies which try to explain how children cope with this ambiguous information. The child might choose one analysis and change it in the course of acquisition (Gibson and Wexler 1994) or have more than one interpretation at the same time and make use of information gathered from the new input during the course of acquisition to find the right and to eliminate other interpretations (Valain 1990). The third strategy is to make no decision until enough information is gathered (Fodor 1998).
A further problem is in the nature of input itself. This was already identified by Chomsky (1981): it only contains positive evidence, which means structures existing in the target language, but does not provide negative evidence. In other words, if a structure does not occur in the input, it can not necessarily be assumed that this structure is ungrammatical. Furthermore, children are not always corrected when making mistakes and corrections are sometimes ineffective as the child can not be sure whether the correction concerns the linguistic form or the content.
For bilingual children, the situation is even more complicated as they are exposed to two sets of input. Several questions and problems arise from this situation: when do the children realise that they are exposed to two languages and have to analyse these inputs in such a manner as to develop two different linguistic systems? How do children deal with ambiguous input when the two languages require different rules in the same grammatical domain?
To answer the first question, De Houwer (1995: 234) points out that research should try to take a different perspective and question why a new-born should expect to be exposed only to one language. It could be possible that the language faculty is able to acquire two or more languages. Research has shown that babies can indeed differentiate between different sounds (Baker 2006: 98) and prefer their mother’s voice to others. Furthermore, they showed reaction to their two languages but when they were exposed to a third unknown one they did not react (Genesee 2003). This proves that they can indeed differentiate between sounds they are exposed to and hence differentiate between different languages. Maneva and Genesee (2002: 383) observed that even in the babbling stage from 10-12 months bilingual children demonstrated language-specific babbling features of each language. Nevertheless, they need to learn to differentiate between the input of their two languages. This process requires the development of awareness in the child that he or she is operating with two linguistic systems. This awareness seems to be there from very early on, as babies already know that different languages are used in different situations with different people (Lanza 1988: 72). Parents and other people in contact with the children can help to support this process. With clarification requests, confirmation checks, failure responses or delays in response they signal to the learner that they do not understand the linguistic form of the learner’s message. The child has to locate and repair the part of conversation which has not been understood and find the right choice of language (Lanza 1988: 72). Lanza (1988) focuses in her study on a bilingual English-Norwegian child, Siri, in Norway from the age of 1;9 to 1;11 . She observed different communication strategies to support language separation. Care-givers either request for the child to repeat the utterance or reformulate the utterance in the requested language. The first strategy, the minimal grasp strategy, places a greater demand to repair on the child. This also gives evidence that the child can differentiate between the two systems. An example by Lanza (1988: 77) presenting a conversation with Siri and her English-speaking mother shows how the mother tries to promote the right language choice.
(1) Situation: Siri has eaten a snack.
Jeg mett. Jeg mett.
(“I full. I full”)
You’re mett? What does Mama say?
Full. That’s a girl!
The mother is stressing the bilingual equivalents mett and full, and gives positive feedback for the right language choice. Furthermore, this example also shows the separation of input according to persons. The “one-parent-one-language strategy” require each of the parents to speak to the child in their (native) language.
The separation by person has been promoted by many researchers (De Houwer 1990, 1995) if this is necessary, but remains controversial as there are other acquisition environments in which the children were also able to separate between their languages (Romaine 1995). In the actual situation total separation of input is nearly impossible. Parents with different native languages talking to each other in one of them will do this in front of the child. (De Houwer 1995: 225) More important is the regular exposure and the same frequency of input of both languages. It has to be kept in mind that bilingual have minimal input requirements to acquire native competence, and that the input of the two languages is already bisected (De Houwer 1995: 227).
The linguistic structure of the input of the two target languages can influence the development of the two systems. There is a very strong correlation between input and child’s output (De Houwer 1997: 159). The linguistic structures the child is exposed to influence the development of the target language. In chapters 3, 4 and 5 there will be a detailed discussion on the ways input promoting different grammatical rules can lead to cross-linguistic influences.
As already mentioned in the preceding discussion, the input of the target language is insufficient to acquire a language without any supporting mechanisms. Furthermore, children acquiring the same language gain the same grammatical competence during the same time period. This leads to the assumption that there have to be underlying principles which allow the child to learn a language. Chomsky (1981) developed the theoretical concept of Universal Grammar (UG), which implies that all languages have the same universal characteristics. UG is genetically determined, and is accessible from birth on. UG can be regarded as a cognitive system with parameters for which different values can be set. The parameters represent grammatical principles which have different values according to the target language. The child has to set parameters via input, which is the language the child hears. First of all, the characteristic has to be identified in input and then to be considered relevant for the setting of a parameter. When the parameter has to be set on a certain value it has to be confirmed by input. (Clahsen 1982: 13-17)
Studies dealing with UG and parameters have mostly been dealing with monolingual FLA, but most researchers assume that bilingual acquisition is basically no different, and that babies appear to be biologically ready to acquire, store and differentiate two languages (Meisel 2004, De Houwer 1990). Nevertheless, some questions and problems arise for BFLA as UG has to be used to acquire two linguistic systems. If it is assumed that parameters cannot be reset and that the two languages have different values for a parameter, it may be asked how the children set different parameter values successively without resetting. An example could be the pro-drop parameter for subjects, which is positive in Italian and negative in German. Italian-German bilinguals have to set two different parameters for their languages. (Müller et al. 2006: 150) As most research dealing with UG concentrates on monolingual FLA, there is little research on how the principles and parameters model of UG applies for two languages. Nevertheless, there are some theoretical assumptions. Köppe (1997: 30) hypothesised that UG can be accessed via different systems, which means that the setting of parameters in one language is independent from having been set in the other already. Time differences in the acquisition of the same feature could even be an indicator for language separation. Meisel (1995: 29) states that “the individual’s ability to set one parameter on contradictory values may very well be a necessary condition for becoming a multilingual.” An example would be the acquisition of functional categories, which for example could occur earlier in one language while in the other the child still uses lexical categories (Gawlitzek-Maiwald and Tracy 1996). It was also shown in Meisel (1994) that the development of both systems runs parallel and that functional categories are integrated at the same time or in one language a little later than the other. This immediately leads to the question of whether this happens by chance or whether the systems interact.
Müller (1994) suggests that parameters might be set on the same value for both languages and be reset afterwards. In this case, parameter resetting would be necessary, which is theoretically problematic. Furthermore, there is no empirical evidence for this theory (Köppe 1997: 30).
Another problem arises when the input is can be interpreted in two different ways and hence, can lead to the setting of two different parameter. Taking the example of the pro-drop parameter in German and Italian again, it could be possible to hear an utterance such as (2) in spoken language although German is a non-drop language.
(2) Du sollst nicht so viele Bücher lesen! – ø Kann so viele Bücher lesen, wie ich will.
(“You should not so many books read! – Can so many books read as I want to”
‘You shouldn’t read so many books! – I can read as many books as I want to’)
According to the example sentence, the child could guess that German is a pro-drop language, but the child would also hear sentences with phonetically realised subjects. Hence, “the data the child is exposed to are sometimes contradictory, s/he could switch parameter values an infinite number of times and as a consequence never settle on the correct value” (Müller 1994: 236). Therefore, the suggestion of a parameter setting constraint was made, which excludes the resetting of parameters during language development (Clahsen 1990).
This lead to other possible solutions of the problem: it was suggested that the child might temporarily work with both values for the parameter. In the course of acquisition, s/he finds out through distribution analyses the positions in which there can be a phonetically non-realised subject (Valian 1990: 115). As these theories do not agree with the theory of UG, the parameter model has been revisited. Haider (1993: 13) puts forward the viewpoint that UG should be seen as a cognitive co-processor and points out that it is activated when it is exposed to data which suits its capacity. The data is then processed fast and subconsciously, which means that the incorporated structure is recognised, stored, retrieved and modified. This implies that the structure of language is not modified by UG, which is not there from birth on, but determined by a programme itself. Language acquisition according to this model is a gradual process as structures are modified, stored and incorporated in the system. In this model parameters are seen as subroutines, which means that it is possible for them to not be simply set on the one or the other value as a parameter, but to be extended (Müller et al. 2006: 36). The two subroutines are built up for the two languages simultaneously or successively. In the successive case, the question arises how the child treats the system which does not yet have a subroutine (Müller et. al. 2006: 38). For monolingual FLA, Roeper (1999) hypothesised that children use a grammar without any analysis of input, which he called Minimal Default Grammar (MDG). An example would be the dropping of verbal arguments. This is revised in the course of acquisition, but not immediately given up. There is a stage when MDG and the new value are alternatively used (Müller et. al. 2006: 38). For BFLA, this means that the child has to define a common system for both subparts of the language systems first. In the course of acquisition it has to test whether the subparts of the systems converge or diverge. Converging subparts are stored as one or two systems as they have the same subroutine. If they are different, they are divergent and are stored separately (Müller et al. 2006: 41).These factors create a special situation in which the two language are acquired and cause distinctive patterns in the course of acquisition in BFLA.
2.3 Concepts and Theories on BFLA
The question of how the two linguistic systems are acquired determined by the factors, which were outlined in the previous chapter, has been of interest for researchers for many years. A prominent theory is the three-stage model by Volterra and Taeschner (1978). They proposed the Single System Hypothesis, which claimed that children start off with one linguistic system which is differentiated in three stages during the course of acquisition. This model was widely accepted until the works of Meisel (1989) and Genesee (1989) initiated reconsideration. They argued for a separation of systems from very early on, and this was proven right in several other studies by different authors in the following years (e.g. De Houwer 1990, 1994, 1995, Meisel 1990a, 1990b, 2001, Köppe 1997, Döpke 1998, 2000a, 2000b). These two positions will be outlined in the following, as they present the two extremes which claim on the one hand complete interdependence, namely a single system, and on the other, complete separation without influence from an early age onwards.
Volterra and Taeschner (1978) described the stages of acquisition a bilingual child goes though from 1;0 to 4;0 and the strategies used by the child during this process. The most important and critical assumption in their study is the claim that in the initial period of acquisition the children develop only one linguistic system before they differentiate the lexical and grammatical systems of their languages. They identified the following three stages in the gradual course of acquisition (Volterra and Taeschner 1978: 304):
I The child has only one lexical system including words from both languages.
II Distinct lexical systems develop, but children still apply the same syntactic rules for both languages.
III The child also differentiates the two languages on the syntactic level, but each language is associated with the parent/person using that language.
At the end of the third stage the child is able to speak the two languages fluently like a monolingual with any person: the “child is truly bilingual” (Volterra and Taeschner 1978: 317).
The data for their study derived from their own longitudinal study of Italian-German bilingual girls, Lisa and Giulia, growing up in Rome, and from Leopold (1970). He kept a diary on the development of his daughter Hildegard who was exposed to German and English up to the age of seven. Lisa and Giulia were observed from 1;5 to 3;6 and 1;2 to 2;6 via monthly recordings of 30 min in alternate language environments and notes taken by the researchers.
In the first stage, the authors assume fusion of the two lexicons on the basis of the word production of their subjects. Their analyses came to this conclusion because first of all, the children’s corpora had little or only very few equivalents. These are words with an identical meaning in both languages. Table 1 (Volterra and Taeschner 1978: 304) shows the words produced in German, English/Italian and words which are closely akin in form and meaning in both languages. An example of an akin word would be “naso”/”Nase” (nose) meaning the same in Italian and in German or “Milch” and “milk” as an example for German and English.
Table 1: Number of words in each child’s production in the first stage
illustration not visible in this excerpt
Not taking the akin words as equivalents, Lisa produced three, Giulia six and Hildegard four (Volterra and Taeschner 1978: 305-307). A second argument for a single lexical system is that the children often do not realise that words exactly correspond to each other and use them in different ways. Lisa uses the Italian word là (there) for thing which are not visible and the German equivalent da for things visible to her (Volterra and Taeschner 1978: 308). They conclude that “the use of one language or another depends upon what the child wants to say and not so much on the language spoken to him [or her]” (Volterra and Taeschner 1978: 309), and hence the child seems to have one lexical system.
In stage two, the children learn more equivalents and use utterances in one language frequently according to the interlocutor, but they apply the same syntactic rules to the two languages. The authors verified this claim by comparing constructions which are different in Italian and in German. Possessive constructions, position of adjectives and negation were examined. Their subject Lisa used non-target language structures for both languages in these domains and acquired the target-specific rules only gradually. They admit that structures found in the bilingual data can also be found in monolinguals’ speech. But on the basis of these structures and the occurrence of mixed utterances, they concluded that Lisa “uses a consistent syntactic system of her own instead of imitating the adult system” (Volterra and Taeschner 1978: 316).
The third stage is characterised by gradual separation of the syntactic systems and they examine the same phenomena as in stage two. The grammar applied is target-language-specific, and as the syntactic rules become apparent, the child also learns to use the language according to the interlocutor’s language.
The three-stage model resulting from this study was widely accepted until the end of the 1980s, when several researchers showed problems in the methodology and the theoretical implications of this study. The articles by Genesee (1989) and Meisel (1989) were the first to address problems of the single-system hypothesis and promoted a separation of systems from early on.
The first point of criticism is the insufficiency of the data gathered to support the fusion of systems, hence “the claim that children initially use only one lexicon is by no means non-controversial” (Meisel 1989: 14). In the first stage, the children are supposed to have a single lexicon because they have only few equivalents. Genesee (1989: 323) and De Houwer (1990: 30-1) state that the data is not sufficient to argue whether the children have a single lexical system. Volterra and Taeschner’s claim is based on the absence of a form and not of the presence which is problematic as there could be more equivalents which were not recorded. This assumption is supported by the fact that the children actually used equivalents. Six equivalent words out of a complete lexicon of 64 words occurred in Lisa’s corpora and 14 out of 73 in Giulia’s. A possible explanation for the absence of equivalents could be that the data was not gathered in different conversational contexts, as some words might be used only in specific situations. For example, if the German-speaking mother takes the child to the kindergarten by car, the child will produce the word “Auto” in this context. The Italian father, on the other hand, might be more involved when the child plays with toy cars and hence, it is likely that only in this context the Italian equivalent “macchina” occurs. Furthermore, the presence of interlocutors from both languages in these different contexts is of importance, as is the use of the same toys, pictures etc. for both languages (De Houwer 1995: 232). That this criticism was legitimate was shown in other studies which found that the children had equivalents in their lexicons. Quay (1995) demonstrated that her Spanish/English subject used equivalents from the beginning of interpretable speech. At the age of 1;5,13 18 out of 47 words were equivalents and formed nine translation pairs. This has been confirmed by Cantone (2007). Her Italian/German bilingual girl Carlotta used 30 German words in an Italian context in stage one (1;8,28-2;3.17 with an MLU ≤ 2). When these cases were analysed in more detail, it became evident that in 17 % of the occurrences of a German word, she produced the Italian equivalent in the same recording. Half of the words occurred in a prior recording session and 7 % of the German words used in Italian contexts were produced in the next recording. Hence, only about 20 % of the words do not seem to have equivalents. An explanation of the missing equivalents could be different input. Some words might be used more frequently in the context of one language, such as for example words concerning childcare, as one parent might be predominantly involved in this task. Hence, a separation of systems takes place from very early on. (Müller et al 2006)
The second point of criticism concerns the interpretation of mixed utterances. Volterra and Taeschner (1978) claimed that there is no regularity within the mixing and regarded this as evidence for a single syntactic system. Genesee (1989) and Meisel (1989) re-examined their claim and found that their empirical evidence and theoretical justification is not sufficient to support their interpretation. If their interpretation is correct, bilingual children would use items from both languages indiscriminately in all contexts (Genesee 1989: 323). Thus, it is necessary to present and analyse the data in the context it occurred in. But Volterra and Taeschner, as well as most proponents of the Single System Hypothesis, did not proceed in this way. Hence, Genesee (1989: 324) argues that “it is impossible to determine whether the children use the repertoire of language items they have acquired to that point in a differentiated way”. The occurrence of mixed utterances might be increased if the interlocutor speaks both of the child’s languages or if the child is exposed frequently to mixed input. Mixing might, furthermore, be a “relief strategy” (Meisel 1989: 14). When children have not acquired appropriate lexical items or grammatical devices in one language, they borrow them from the other language. This strategy can be compared with monolingual children’s overgeneralization when they expand the meaning of a word they already know to express another not yet known. Bilinguals can overgeneralise inter-lingually as well as intra-lingually when they borrow from the other language (Genesee 1989: 325). This hypothesis is supported by the decreasing of mixing in the course of language acquisition. As more words and grammatical structures are learned, the need to take structures and words from the other language to express oneself is diminished. Whereas Volterra and Taeschner regard mixing as evidence for a single system, Meisel (1989) sees the ability to mix the two languages as part of the pragmatic competence of the bilingual. He proposed a terminology which defines this mixing as the speaker’s ability to coordinate the languages in accordance with the grammatical constrains of both languages, and calls it code-switching. The indiscriminate combination of elements, on the other hand, is called language-mixing. Meisel (1994) observed code-switching in young bilingual children and came to the conclusion that it already demonstrated forms and functions of adult code-switching. Consequently, these findings strongly support the assumption that the two linguistic systems are separated from early on.
The study by Volterra and Taeschner (1978) also has some logical and theoretical deficiencies. First of all, the definition of the stages is surprisingly vague: there are no independent criteria given such as age or MLU, or any feature which is not part of the definition itself (Meisel 1989: 15). Another problem is that the phenomena which characterise stage one and two, namely mixing of lexical items or of grammatical structures, do still occur in the later stages although the lexicons and syntactic systems are supposed to be separated. Hence, this model is disproved and strong evidence is provided for an early separation of the two systems.
 Participants of this study used the "one person one language" strategy