Master Thesis General Linguistics, Graduate School of Humanities, University of Amsterdam
I Love You!
Student: Katya ButsStudent number: 10079068 Supervisor: Dr. R. Pfau
Second reader: Prof. Dr. B. van den Bogaerde
Date: June 8, 2015
imodal ilinguals
Acknowledgements
I would like to express my appreciation to the numerous persons who helped me in the preparation, writing and finishing of this thesis. First of all, much appreciation goes to my supervisor Roland Pfau for his expertise, inspiration, support and willingness to assist me solving the problems I encountered, both in the planning and in the process of writing this thesis. I really appreciate the trust and patience he showed for the extra time I needed to finish my thesis.
Furthermore, I would like to thank the NSDSK (Nederlandse Stichting voor het Dove en Slechthorende Kind, Dutch foundation for the Deaf and hard-of-hearing child) for the opportunity to watch and use some of their recordings for further analysis. Special thanks should go to Evelien Dirks who inspired and encouraged me to write a thesis about the linguistic aspects of these recordings, which initially were made for a research project investigating the social interaction between a parent and a child.
Abstract
In terms of production, unimodal bilinguals are able to switch between spoken or sign languages (i.e. code-switching). However, bimodal bilinguals can not only mix their languages but they can use them simultaneously in distinct modalities (i.e. code-blending). This possibility allows for simultaneous production of two languages with an unique way of language mixing as a result. Previous studies show that the simultaneous use of signs and speech mostly occurs in deaf and hearing children’s output and input in deaf families. In addition, bimodal bilinguals even seem to prefer to use code-blending over code-switching.
The present study investigates the production of code-blending from two hearing children
(age 5;9 and 5;11) growing up in families with at least one deaf family member, and their mothers. Therefore, these children are native users of Sign Language of the Netherlands (Nederlandse Gebarentaal, NGT) and Dutch. The main research questions are the following: what types of code-blending are produced by approximately six-year old hearing children of deaf parents? And do the blending patterns produced by the children reflect the code-blending patterns found in the output of the mothers directed towards the children?
Analysis of two recordings of spontaneous speech from both children interacting with their mothers, shows that the children frequently produced code-blends. Results also show a certain amount of all classified types of code-blending (as described in this study) produced by both mothers and children. However, where analysis of the mother’s utterances shows similar code-blending patterns for both mothers, the patterns found in the utterances of the children differ from each other.
Contents
1. Introduction……….. p. 4 2. Theoretical Background……… p. 6 2.1 Typology of code-mixing………. p. 6 2.1.1 Classification of code-switching types……… p. 6 2.1.2 Bimodal bilinguals and code-blending……… p. 8 2.1.3 The role of code-mixing in language input to children……..…... p. 11 2.1.4 Determining the ‘matrix’ language…………..……… p. 12 2.2 Mouthing and word or sign classes ……….……. p. 13 2.2.1 Mouthing……….. p. 14 2.2.2 Word or sign classes in code-blending………..………. p. 15 3. Method……… p. 16 3.1 Procedure of data collection.………. p. 16 3.2 Subjects……….. p. 16 3.3 Annotation and analysis………. p. 17 4. Results………..……….. p. 20 4.1 Comparison of types of code-blending used by the children ……… p. 20 4.2 Quantitative results of code-blending types produced by the
mothers………. p. 25 5. Discussion………..……….. p. 27 5.1 Analysis of code-blending used by the children……… p. 27 5.2 Analysis of code-blending: a comparison of mother and child………….. p. 28
5.2.1 Language input from the mothers compared to language
output of the children………. p. 28 5.2.2 Challenges in determining the matrix language………. p. 31 5.2.3 Analysis of word or sign classes produced by mothers and
children in blendings……….. p. 32 6. Conclusions..……… p. 35 References………. p. 37 Appendices Appendix A: Transcript 1...……… p. 39 Appendix B: Transcript 2...………. p. 56
1. Introduction
When bilinguals communicate with each other, they typically mix their languages, that is, they use elements or structures from their two languages (Muysken 2000). In general, language acquisition studies of hearing children growing up with bilingual input, find that if the parent(s) mix their language, the children are influenced by this mixed input. For instance, children acquiring two (or more) languages, tend to mix these languages, often from the very beginning (Van den Bogaerde & Baker 2005). In deaf families and in families with both hearing and deaf members, both the sign language and the spoken language can be used, resulting in bimodal bilingual family members, who are fluent in a spoken and a sign language.
This is an interesting phenomenon that will be explored further in this thesis by looking at the language output of two hearing children growing up in deaf families, in particular the language mixing that occurs in the language output of the children. In both families, Sign Language of the Netherlands (Nederlandse Gebarentaal, NGT) and Dutch are used. Van den Bogaerde & Baker (2005) describe a similar type of situation, focusing on children up to three years of age. The novel contribution of the present study is that the research will focus on a population of older hearing children, namely two girls of almost six years of age, while adopting the same classifications as the study by Van den Bogaerde & Baker (2005). Also, the more fine-grained classifications used by Donati & Branchini (2012) are applied in addition to Van den Bogaerde & Baker’s (2005) classification to give a more detailed picture. Furthermore, the matrix language of the utterances from both the mothers, as well as from the children, will be compared to determine if one of the languages is used predominantly. In addition, this study will not only focus on language mixing patterns, but will also quantify word or sign classes used in the language mixing in the output of both mothers and children.
The research questions for the present study can be formulated as follows:
1. What types of code-blending are produced by six-year old hearing children of deaf parents?
2. Do the code-blending patterns produced by the children reflect the code-blending patterns found in the output of the mothers directed towards the children?
Several patterns might occur in the results:
Analysis of the children’s utterances shows similar patterns for both children and these patterns are comparable to the patterns found in the utterances of the mothers.
Analysis of the children’s utterances shows similar patterns for both children, these patterns are different from the patterns found in the utterances of the mothers. This would mean the children adapt or use the input provided by the mothers in their own way.
Analysis of the mother’s utterances shows similar patterns for both mothers, however, the patterns found in the utterances of the children differ from each other. First, we present the relevant theoretical background information (chapter 2), attending also to the relevant terminology of code-switching, code-blending and matrix languages. Also, a brief outline is provided about sign language use and bimodal bilinguals. Chapter 3 describes the methodology of the present study. The results are presented in chapter 4, followed by a discussion, in particular a comparison of the children and their mothers, in chapter 5. Chapter 6 concludes the study and offers an answer to the research questions.
2. Theoretical background
This chapter provides relevant background information for the present study, starting with a description of the typology of code-mixing in 2.1. In this section, we introduce the classification of code-switching types in spoken languages (2.1.1) and we sketch characteristics of sign language use in bimodal bilinguals (2.1.2), thus providing an overview of the use of code-blending types in sign languages. In 2.1.3 literature that concerns the role of code-mixing in language input to children is discussed, and in 2.1.4 the terminology of determining a matrix language is considered. Subsequently, in 2.2, the role of both mouthing as well as word- or sign classes in code-mixing is discussed.
2.1 Typology of code-mixing
In many bilingual communities bilingual speech can be found. When an individual is competent in two (or more) languages, s/he has the possibility to access both languages in the same utterance, even though the two languages involved in such a situation might display different word orders. According to Muysken (2000), bilingual speech involves extensive code-switching (which is a type of code-mixing), in which an individual switches from one language to another within one utterance or sentence. This could result in a bilingual sequence consisting for example of ABABAB, where A and B represent two different languages. Code switching is a very widespread and systematic feature of the production of any bilingual. It can be defined as “the juxtaposition within the same speech exchange of passages of speech belonging to two different grammatical systems or
subsystems” (Gumperz 1982: 59, in: Donati & Branchini 2012). Switching can occur at
different places in speech, either at clause boundaries, for example ‘Yo quiero agua because I’m thirsty’ or within the boundaries of a clause or sentence, for example ‘Yo quiero water’ (Donati & Branchini 2012: 94). The following section (2.1.1) will provide an outline of the classification of code-switching types according to Muysken (2000). In 2.1.2 this classification is related to the classification of the code-blending (which is also a type of code-mixing) types discussed in Donati & Branchini (2012).
2.1.1 Classification of code-switching types
Deuchar, Muysken & Wang (2007) describe the development of precise criteria to establish profiles for bilingual speech, following the typology of insertion, alternation and congruent lexicalisation developed in Muysken (2000). Adopting these criteria, they demonstrate that bilingual speech, just as monolingual speech, shows variation, but that this variation is patterned or structured, not random. Muysken (2000) characterised the phenomenon of code-switching in terms of three competing bilingual speech strategies:
The insertion of material (most often a word or a constituent) from one language
The alternation between stretches of words in different languages.
The congruent lexicalisation of a shared language structure with words from
different languages.
One of these patterns usually dominates, though this will not necessarily result in the exclusion of other patterns. In the insertion pattern, one language determines the overall structure, into which constituents from the other language are inserted. This can be illustrated by example (1), of a Swahili-English sample (with English marked in bold) from Myers-Scotton (1993: 86, in Deuchar, Muysken & Wang 2007):
(1) a-na-ku-l-a plate m-bili z-a murram
3SG-PRS-NFIN-eat-IND CLM 10-two CLM 10-of maize
‘He eats two plates of maize.’
In this example, the Swahili word order is utilized, including the phrase in which the English noun ‘plate’ is used (plate m-bili ‘two plates’). Also, all the inflectional morphology in the utterance is from Swahili.
In the alternation pattern, both languages (English and Dutch in example 2) occur alternately, each following their own structure, as illustrated in example (2) from Van den Bogaerde & Baker (2005:153):
(2) I want dat je mij zoent
I want that you me kiss ‘I want you to give me a kiss.’
The third type of code-switching is congruent lexicalisation, following the definition that “the grammatical structure is shared by languages A and B, and words from both languages A and B are inserted more or less randomly” (Muysken 2000: 8). This type of code-switching is most often present in mixing between dialects and standard languages, which are close to each other in structure. This type can be illustrated by example (3) from the Ottersum dialect (in bold) and standard Dutch, in this example the fragments from each variety apparently do not form grammatically coherent chunks (Muysken 2000: 130, citing Giesbers 1989: 147):
(3) ja maar bij ouwe mensen komt dat gauw-er tot stilstand
yes but with older people comes that quick-er to-a halt
als bij jonge mense wa
than with younger people eh
These three code-switching types have also been described for sign languages (Donati & Branchini 2012), and this is considered in the following section (2.1.2).
2.1.2 Bimodal bilinguals and code-blending
After several decades of linguistic investigation, sign languages have been shown to be natural languages, fully capable of expressing any thought or emotion and complete with a rich lexicon and complex grammar. Also, acquisition studies have shown that deaf or hearing children acquiring a sign language from their deaf parents, learn that language from birth in a conventional and natural way, attaining sign language acquisition milestones on a timeline similar to that of a child learning a spoken language. An individual who is a native signer should not be considered language-impaired or language-delayed. Instead, this individual probably acquires two or more languages (including a sign- and a spoken language) and therefore should be seen as a bilingual (Singleton & Tittle 2000). Bimodals are bilingual individuals for whom the two languages involved belong to two different modalities. Therefore, bimodal bilinguals provide an exceptional circumstance in which mixed utterances are either perceived through the acoustic channel and produced through the vocal track, or perceived through the visual channel and produced by movements of hands and body. So, while the physical constraints of the vocal tract force bilinguals of two spoken languages to produce mixed utterances with the two languages sequentially alternating, the physical independence of the two linguistic channels exploited by bimodals allows them not only to code-switch sequentially, but also to simultaneously mix bits of the two languages. Therefore, Emmorey et al. (2005) established the distinction between code-switching and code-blending. In their definition, code-code-switching between sign and spoken language is to stop talking and switch to signing. However, bimodals do not need to stop talking in order to sign, or vice-versa, but may speak and sign simultaneously. This so-called code-blending can be defined as signs produced simultaneously with spoken words (Donati & Branchini 2012). This term will be used in the present paper where relevant to refer to the simultaneous mixing of signs and words.
Although the bimodal population includes both deaf and hearing users, linguistic and socio-linguistic research has focused on hearing children and adults born in deaf families, so-called CODA’s (Children of Deaf Adults). CODA’s acquire both their spoken and sign language(s) naturally, as any native bilingual does when exposed to two languages from birth. Hearing bimodals acquire the two systems in parallel and spontaneously without any explicit instruction as they have natural access to both languages, as opposed to deaf users of a spoken language and a signed language, who do not have natural access to the spoken language modality. As a result, CODA’s are more likely to be balanced bilinguals, who are exposed to sign language at home and to spoken language in the extended family situations, on various social occasions and at school. In contrast, deaf bilinguals may very well be competent users of the spoken language, but they do not have spontaneous access to it due
to their hearing impairment. Therefore, the acquisition of the spoken language is likely to be rather delayed in time and possibly requires long lasting and specific speech training (Van den Bogaerde & Baker, 2005).
Donati & Branchini (2012) describe code-switching and code-blending patterns that can be expected in the language use of bimodal bilinguals. Previous studies consistently confirm that code-blending is a massive phenomenon among bimodals and even that it is by large the preferred option for communication. Emmorey et al. (2005) report that the subjects investigated (CODA’s from the United States) stop talking in order to sign (code-switching) in only 5% of the recorded mixed productions, whereas in 95% of the remaining
utterances signs and words are produced simultaneously.
The fivecode-switching and code-blending patterns that Donati & Branchini (2012)
describe will be used in the present study and can be explained as follows: 1. (Lexical) Insertion
This type of blending includes an autonomous and complete utterance in just one modality, supported by a few words or signs in the other modality, that do not add any extra meaning to the global utterance. The words or signs "duplicating" the dominant utterance do not need to be synchronized with their equivalent in the dominant channel. This type of blending can be related to the insertion pattern described by Muysken (2000) and is
illustrated in (4) below, where the sign GIVE in Italian Sign Language (LIS, Lingua dei Segni
Italiana) is occurring synchronized to the corresponding Italian verb, since both words are articulated simultaneously (Donati & Branchini 2012: 104):
(4) Italian: La strega dà la mela a Biancaneve
The witch give.3SG the apple to Snow White
LIS: GIVE-CL
‘The witch gives the apple to Snow White.’ 2. Alternation
Sometimes, this type is also referred to as a form of code-switching, since there is an interruption of signing in order to speak or vice versa. This is illustrated by example (16) of an utterance produced in the present study (section 4.1, p.22). This type can be related to the alternation pattern described by Muysken (2000), where both spoken languages occur alternately, each following their own structure, for an example see English and Dutch in (2).
3. Independent Blending
This type of blending involves the simultaneous production of two independent and apparently autonomous monolingual utterances. The two produced sentences may be slightly different in meaning, as illustrated in (5) where the utterance is semantically
(5) Italian: Lavora a Rimini
work.PRS.3SG in Rimini
LIS: WORK LOC
‘He works there, in Rimini.’
This type of blending is sometimes related to the insertion pattern described by Muysken (2000). However, it is not the same because the insertion of material (most often a word or a constituent) from one spoken language into an utterance in another spoken language, is not possible simultaneously. Therefore, it is not possible to produce two sentences that may be slightly different in meaning, as is possible in Independent Blending.
4. Congruent Lexicalisation
A type of blending where two blended strings happen to display the same word order in the two languages according to their specific grammars, therefore there is a complete and natural matching between the words and the signs produced simultaneously. An example of this phenomenon in Italian and LIS is provided in (6) (Donati & Branchini 2012:107).
(6) Italian: Lei sa tutto
she know.3SG everything
LIS: INDEX3 KNOW ALL
‘She knows everything.’
This type is similar to the congruent lexicalisation pattern described by Muysken (2000). This is a description of shared language structures with words from different spoken languages that occur in a sentence alternately. However, the definition that “words from both languages A and B are inserted more or less randomly” (Muysken 2000: 8), does not specifically apply to blended Congruent Lexicalization of a spoken and a sign language.
5. Blended Blending
This type of blending consists of a mixed utterance of which the constituents are distributed across the two channels. Crucially, the utterance is complete and meaningful only if the fragments distributed in the two channels are put together in a unique, blended utterance. The two modalities may also include some reduplicated elements. This type of blending is illustrated in example (7) (Donati & Branchini 2012:110), showing an utterance in which Italian provides the indirect object while the utterance in LIS provides the (postverbal) subject, and the verb is produced in both channels.
(7) Italian: Parla con Biancaneve
talk.PRS.3SG with Snow White
LIS: TALK HUNTER
An overview of all types of blending used in this research can be found in Table 1.
Table 1:Types of blending, according to theclassification by Donati & Branchini (2012).
2.1.3 The role of code-mixing in language input to children
Various factors determine the differences in types of code-blending used by children. Amongst other things, these factors appear to be linked to the language ability of, the language input to, and the language choice of a child (Baker & Van den Bogaerde 2008). The present study focusses on the language input directed to the CODA’s and the language choice of these children. Therefore, this section discusses results from previous studies that have focused on the language input directed to children growing up in families with (a) deaf member(s). In sections 2.1.4 and 2.2, research focusing on language choice will be discussed.
Research in the United States has revealed that deaf adults mix spoken English and American Sign Language (ASL) amongst themselves (Lucas & Valli 1992). According to the same study, deaf adults also produce mixed utterances in interactions with hearing people who are bilingual in ASL and English. Therefore, it can be expected that deaf parents of hearing children also tend to mix the language directed towards the child. This is indeed confirmed by the results of Mallory, Zingle and Schein (1993: in Baker & Van den Bogaerde 2008), showing that code-blending in ASL and English also commonly occurs in interactions between deaf parents and their hearing children. Furthermore, Petitto et al. (2001) studied young CODA’s acquiring French and Quebec Sign Language (LSQ). The children in this study all were between 0;10 and 4;3 years of age. The children and their parents both produced code-blended utterances. Based on their finding, the authors argue that the children participating in this study are not delayed in the acquisition of their first word and word combinations in French and equivalents in LSQ because all children achieved their early
Types of blending Definition Relation to Muysken’s
(2000) categories 1. (Lexical)
Insertion Autonomous, complete utterance in one modality, supported by a few words or signs in the other modality.
Relation with:
Insertion 2. Alternation Interruption of signing in order to speak or vice
versa. Relation with: Alternation pattern 3. Independent
Blending Simultaneous production of two independent, autonomous monolingual utterances. The two sentences may be slightly different in meaning.
None
4. Congruent
Lexicalisation Two blended strings displaying the same word order in two languages according to their specific grammars. Complete matching between
simultaneously produced words and signs.
Relation with:
Congruent lexicalisation 5. Blended
Blending Mixed utterance of constituents distributed across two modalities. This only is meaningful if the fragments distributed in the two channels are put together in a blended utterance.
linguistic milestones in both their languages at the same time (and similarly to monolinguals), and they also demonstrated sensitivity to the interlocutors’ language by adapting their language choices. The children did use code-blending to varying degrees, and some persisted in using a language that was not the primary language of the addressee, but the propensity to use both languages at the same time was directly related to their parents mixing rates, in combination with their own developing language preference.
Also, previous research by Van den Bogaerde (2000), showed that in the acquisition of NGT and spoken Dutch, code-blending is frequently observed in interactions between deaf parents and their children between 1;0 and 3;0 years of age.
Given that these previous studies showed that the amount of code-blending in the language input of the children was of influence on the children’s output, it is interesting to relate the language input to the children’s language output in this study as well. To do this, the amount of code-blending occurring in the input from the mothers directed to the children is quantified and the results are related to the amount of code-blending observed in the output of the children. Furthermore, it is interesting to compare the matrix language of the utterances from both the mothers, as well as from the children, to determine if one of the languages is used predominantly (see section 3.2). To that end, we will establish criteria for determining the matrix language in the next section.
2.1.4 Determining the ‘matrix’ language
The code-blended utterances described in 2.1.2 can be divided into four sub-classifications in order to determine the matrix-language. To accomplish this, the idea of a semantic base can be used. For example, when a proposition is expressed fully in spoken Dutch with only semantically congruent signs, the code-blended utterance can be classified as Dutch Matrix Language. For this classification, only the proposition is used, morpho-syntactic criteria are not used because the child only has emerging competence. Therefore, the use of morphological elements to determine the matrix language, as is done in Myers-Scotton (1993, in Van den Bogaerde & Baker 2005), could lead to an incorrect classification, since these elements are in the process of being acquired (Van den Bogaerde & Baker 2005). Van den Bogaerde & Baker (2005) propose the following matrix language classification:
1. Code-blended, Dutch Matrix Language
Type of utterance in which each occurring sign is semantically congruent with one (spoken) word, therefore, the proposition is expressed entirely in words. Expressed signs do not contribute additional meaning to the utterance. In accordance with the classification of Donati & Branchini (2012), this classification type could be labelled as a form of (Lexical) Insertion or Alternation. Example (1) would thus be classified as ‘Italian Matrix Language’.
2. Code-blended, NGT Matrix Language
Type of utterance in which each occurring (spoken) word is semantically congruent with one sign, therefore, the proposition entirely expressed in signs, where words do not contribute
additional meaning to the utterance. In accordance with the classification of Donati & Branchini (2012), this classification type could also be labelled as a form of (Lexical) Insertion or Alternation.
3. Code-blended, Full
Utterances in which the full proposition is expressed in both modalities. The utterances do not have to be complete structurally, in either NGT or Dutch. In accordance with the classification of Donati & Branchini (2012), this classification type can be labelled as Independent Blending.
4. Code-blended, Mixed
Type of utterance in which both the signs and spoken words are necessary to make up the full proposition. In accordance with the classification of Donati & Branchini (2012), this classification type can be related to Alternation and Blended Blending.
Utterances of the type Congruent Lexicalisation cannot display a matrix language because in this type of blending, two blended strings have to display the same word order in two languages according to their specific grammars, which results in a complete matching between simultaneously produced words and signs. Therefore, the blended languages are uttered in an identical manner without one language being predominant over the other. The present study will focus on the types classified as number 1 (Dutch Matrix Language) and number 2 (NGT Matrix Language) and a third sentence type which displays no clear
matrix language (also known as ø Matrix). In utterances classified as ø Matrix no matrix
language could be determined because both the expressed signs and spoken words
contribute additional meaning to the utterance.
Van den Bogaerde & Nortier (2006) looked at the language production of a six year old hearing boy and his deaf mother in an informal conversation. They tried to establish the matrix language used by the mother and the child. The results of their research show a clear matrix language used by the mother, namely NGT. However, in the language production of the child they found a considerable amount of utterances where no matrix language could be determined. They propose that the age of the child or his hearing status might have influenced the production of the (no clear) matrix language.
2.2 Mouthing and word or sign classes
When signers of NGT sign with each other, they often combine their signing with silent articulations of lexical items from spoken Dutch, or ‘mouthings’. Mouthings are a paramount feature in NGT conversations and can be analysed as instances of code-blending (Bank 2014). The blending classifications introduced in the section 2.1.2 are therefore instances of code-blending. In section 2.2.1, a description of several types of mouthing is given, followed by a discussion of previous research results found for mouthing occurring in
code-blended utterances. Also, a relation is observed between word class and mouthing. This is discussed in 2.2.2.
2.2.1 Mouthing
Mouth actions or mouth movements can be divided into two subgroups: mouth gestures and mouthings, according to Crasborn et al. (2008). The most important difference between these subgroups is that mouthings are derived from a spoken language, as they represent (part of) a spoken word, while mouth gestures are not derived from spoken language. Bank (2014) states that mouthings are not specified in the NGT lexicon, but are always instances of code-blending. Therefore, the distinction between mouth gestures and mouthings is relevant to the present study.
Schermer et al. (1991) distinguish between three types of mouthings. The first type are mouthings produced without any signing or hand movements. This type of mouthing occurs in alternating sentences, where both languages occur alternately, each following their own structure (see 2.1.2). The second type are mouthings that are combined with a
sign and form a word. Examples of this kind of mouthing are the signs BROER (‘brother’) and
MAN (‘man’) in NGT combined with the mouthings broer and man, respectively. The third
type are mouthings that represent part of a word that the sign refers to. Examples of this
kind of mouthings in NGT are ontwik and rodam combined with the signs ONTWIKKELEN (‘to
develop’) and ROTTERDAM.
In the context of language input from deaf mothers to hearing children, results in Baker & Van den Bogaerde (2008) show that mouthings occur frequently. According to the authors, the factors that determine the occurrence of these mouthings are clearly related to the hearing status of a child and probably also to the child’s knowledge of the spoken language. Other research observed for NGT as well as for some other sign languages, that linguistic factors also play a role in the occurrence of mouthings (see papers in Boyes Braem & Sutton-Spence 2001). For example, in adult signers, mouthings have been found to occur more often with nouns than with verbs. In contrast, mouth gestures (mouthed non-words) more frequently accompany verbs as part of their morphology. However, for hearing bilingual adults using ASL and English the study of Emmorey et al (2005) found opposite results. They observed that for these subjects, verbs were the largest category occurring in blended utterances, that is, verb signs were more likely to be accompanied by mouthings. According to the authors, this result indicates that verbs are easily code-blended in sign-speech combinations. This could be explained by the fact that the two morphological systems do not impede each other, that is, spoken tense inflections can remain while signed verbs are produced combined with their own morphology. The authors also state that nouns are possibly less code-blended because subjects can be omitted in ASL. The present study carries out a further analysis at the lexical level of the types of word/sign classes with which
code-blending occurred, in order to investigate the role of the linguistic factors influencing the code-blending occurring in the utterances of Dutch and NGT.
2.2.2 Word or sign classes in code-blending
Linguistic factors could be involved in the occurrence of code-blending (see 2.2.1). Muysken (2004:153) discusses code-mixing in spoken languages and provides a hierarchy in the categories used in the mixing of spoken languages (> meaning ‘more than’): nouns > adjectives > adverbs > verbs >adpositions > conjunctions.
Baker & Van den Bogaerde (2008) examined the distribution of word/sign types in code-blending produced by hearing CODA’s and deaf children, including only the elements that had the same semantic content in sign and in word. In the analysis, they looked at the production of the following five categories: verbs, nouns, pronouns, adjectives/adverbs and question words. They found that nouns are produced most often as code-blends. This is comparable with the results that Muysken (2004) found for code-switching in spoken languages. However, the category adjectives/adverbs occurred less frequently in the results of Baker & Van den Bogaerde (2008) than in the results found for spoken languages (Muysken 2004). Also, verbs are found to occur quite often in the code-blends in Baker & Van den Bogaerde (2008). However, in contrast with findings from Emmorey et al. (2005), verbs do not emerge as the word class most strongly associated with code-blending in the study of Baker & Van den Bogaerde (2008).
In their results, Baker & Van den Bogaerde (2008) omitted the category question words because these were produced relatively infrequently. They provide the following hierarchy in the categories used in code-blending of Dutch and NGT produced by hearing children: Noun > Verb > Pronoun > Adj/Adv. As for the input to hearing children, they found the following hierarchy: Noun > Verb > Adj/Adv > Pronoun.
The differences found in the results of Emmorey et al. (2005) and Baker & Van den Bogaerde (2008) might be due to the age discrepancy in both studies. In the study of Emmorey et al. (2005), participants were between 22–53 years old, with a mean age of 34. However, Baker & Van den Bogaerde (2008) studied children between 2;11 and 3;0 years of age, with one participant also being studied at 5;09 years of age. This could imply that there is an age-related difference in the hierarchy of code-blending types used by bimodal bilinguals. The participants in the present study are a little older than the participants in Baker & Van den Bogaerde (2008), namely 5;11 and 5;9 years of age, and therefore it might well be the case that they produce code-blends according to a hierarchy that is a mix between the results found for older bimodal bilinguals (Emmorey et al. 2005) and younger bimodal bilinguals (Baker & Van den Bogaerde 2008). The words/signs used in code-blending will be further discussed in 5.2.3, after a description of the methodology is given in the next chapter and the results of the present study are presented in chapter 4.
3. Method
This chapter describes the methodology of the present study, starting with a description of the procedure of data collection (3.1) and the subjects participating in the study (3.2). To investigate quantitative and qualitative properties of code-blends produced by the subjects, a 5;9 and a 5;11 year old child, and to identify the blend-types (see Table 1) that were produced, two recordings have been transcribed and analysed for the children’s spoken and signed utterances. This procedure of annotation and analysis will be further discussed in 3.3.
3.1 Procedure of data collection
The data discussed in this section is part of a research project investigating the social interaction between a parent and a child conducted by the NSDSK (Nederlandse Stichting voor het Dove en Slechthorende Kind, Dutch foundation for the Deaf and hard-of-hearing child), an Amsterdam-based institute, which is specialized in issues related to language and hearing. The recordings used for analysis in chapter 4 are from this project, in which recordings were made of a deaf parent and a hearing child, playing with several toys in sessions that lasted for approximately twenty minutes. These two particular recordings were chosen for analysis because they were of good quality. Other recordings were problematic because the frame would not always show both the child’s and the mother’s face and hands. Therefore, it was not always clear in the other recordings what was being said or signed. Also, in some recordings the child was very shy and intimidated by either the camera or the researcher present in the room. Consequently, these children were reluctant to communicate with the mother and some of them tried to hide their hands and faces from the camera when they did respond to the mother. As a result, these recordings were not suitable for analysis of the language utterances.
3.2 Subjects
In the present study, the recording SLM 01.MTS and CODAEAS_180407 are used for analysis. In the recordings, which have a length of respectively 19:01 and 13:33 minutes, a deaf mother and her hearing daughter can be seen. The child is playing with the mother and is naming objects (farm animals) or telling stories, for instance about school. During the recordings, the topic of conversation changes several times (see appendices 1 and 2 for the full transcript of the spoken and signed utterances in both conversations; except for mouthings, non-manual features are not included in the transcription). The topics of conversation are similar in both recordings. In the present study the code-blended utterances of both the children and their mothers are analysed for word or sign classes produced in the blendings and this might be influenced by the topic of conversation. Therefore, we decided to analyze both of the recordings completely and present the results in percentages so no lexical content is lost as a result of cutting recording SLM 01.MTS to
correct for the difference in time of both recordings. The difference in total time of both recordings might have influenced the results because child 1 had more to interact with her mother. However, the total amount of utterances shows that child 1 did not produce more utterances (69 in total, whereas child 2 produced 84 utterances in total).
The children are 5;11 and 5;9 years of age and participated in the control group of the original study (entitled: emotionele beschikbaarheid in de ouder-kind interactie van dove
ouders en horende kinderen, emotional availability in parent-child interactions of deaf
parents and hearing children) which is a still ongoing research project performed by the
NSDSK. Both children were selected to participate in the control group of the original study because the family has no social or language support from the NSDSK and the children are not known to have developmental problems. Table 2 provides the relevant information on the subjects participating in this study.
Table 2: Information on the subjects participating in this study
3.3 Annotation and analysis
The children’s spoken/mouthed and signed utterances are transcribed and analysed for code-blending phenomena. For this purpose, utterances including a combination of signs and words are classified according to the five classifications of code-blending (see Table 1). Phonation was not a criterion for an element to be considered a spoken word, the words could be mouthed, whispered or spoken with voice. Utterances were counted in accordance with the following definition: an utterance is a string of signs and/or words that form a unit on a syntactic, semantic and pragmatic level (Baker et al. 2005). For signed utterances, the utterance boundary is further established by noting when the hands go into a rest position, for instance: on the lap, on an object or person, or in front of the body (Bos et al. 1988, in: Baker & Van den Bogaerde 2008). For spoken utterances, the utterance boundary is further indicated by pause-length and intonation. Also, turn-taking of the addressee is an indicator of an utterance boundary in both languages.
The utterances from the children are interesting in themselves becausechildren are
reported to be likely to produce mixed utterances in any language pair (Meisel 1989, in Donati & Branchini 2012). Also, for every utterance, we will determine whether one language is the matrix language in the code-blended utterance, or whether there is no specific matrix language. See, for instance, examples (8) and (9) below. In (8), no matrix
Participant Hearing status Age at filming Length of recording Mother Father
Child 1 Hearing 5;11 19:01 minutes Deaf Hearing
status unknown
Child 2 Hearing 5;9 13:33 minutes Deaf Hearing
status unknown
language can be determined because there are signed as well as spoken words and the grammar does not specifically follow the grammar of either signed NGT or spoken Dutch. As discussed in 2.1.4, this type of matrix language is classified as ø Matrix in the present study. (8) Spoken water auw voorzichtig
Signed WATER headnod PAIN CAREFUL
‘That water hurts ouch, be careful.’
However, in example (9) only the word hebben (‘have’) is both uttered in spoken Dutch and NGT. The complete sentence is in Dutch, uttered according to the Dutch grammar. Therefore, this utterance can be classified as NL (Dutch) matrix.
(9) Spoken wil je chocola hebben
Signed HAVE
‘Do you want chocolate?’
The code-blending types and use of matrix language found for both children are compared to the code-blending types and use of matrix language found for the mothers to allow for an analysis of the relation between the language input to the children and the language output produced by the children. Furthermore, the code-blended utterances of both the children
and their mothers are analysed for word or sign classes produced in the blendings. In
example (8), the blended elements were classified as Noun (for the word water) and Adj/Adv (for the word voorzichtig), and in example (9) as Verb (for the word hebben). As can be noted, both examples (8) and (9) also show an Alternation pattern (discussed in 2.1.2). However, this type can also be referred to as a form of code-switching, and because of the ongoing discussion about the status of an alternating sentence as a blending or code-switching phenomena, in the present study, we only counted the amount of Alternation occurring in the utterances. The word or sign classes in these alternating utterances were not included in the count and analysis of the word or sign classes produced in the code-blended utterances. Also, the counting of Blended-Blended utterances was more complex because the spoken and the signed element could often be classified as different word classes, therefore the spoken and signed elements of all Blended-Blended utterances were counted separately (for example see utterances C1.42 in appendix A and C2.06 in appendix B).
We know from previous research that it may occasionally be hard to interpret or classify utterances, especially when there is frequent use of gestures that can also be NGT signs. For further research, it might be interesting to study this phenomenon. Within the limited scope of this thesis, possible gestures were counted as signs or, when this was problematic, decisions were taken based on the context or code-blend, see example (10).
This example of Lexical Insertion shows an interesting combination of spoken language combined with a sign from NGT which is also known as a gesture in spoken Dutch. There were no other NGT signs accompanying the spoken utterance of the child.
(10) Spoken lekker tis zoete peper
Signed NICE
‘Yummy, it’s sweet pepper.’
After some consideration, this utterance was classified as a Lexical Insertion blending type,
because the gesture ‘lekker’ is also known as the sign NICE (LEKKER) in NGT. However, given
that the ‘signed’ part resembles gesture more than NGT, there might be other interpretations which would lead to an alternative outcome and classification.
With Blended-Blended utterances, the blending could contain a gesture and a spoken word (both different in meaning). The gesture was then counted as a NGT sign because both the gesture and spoken word added meaning to the utterance (for example see C2.67 in appendix 2).
4. Results
This section provides an overview of the results from the research described in chapter 3. Also, some interesting examples from the transcriptions are given. For the complete transcriptions see appendices A and B. The focus in this chapter will be on the children (examples and results), and at the end, we will also briefly provide the results found for the mothers, which will then be further discussed in 5.2 and 5.3.
4.1 Comparison of types of code-blending used by the children
The first child (C1) produced 69 utterances in total in 19:01 minutes. Of these 69 utterances, 32 utterances (46% of all utterances) were mixed, the other utterances were either spoken or signed. The second child (C2) produced 84 utterances in total in 13:33 minutes. Of these utterances, 69 utterances (82% of all utterances) were mixed. The other utterances were either spoken or signed. Tables 3 and 4 provide an overview of the distribution of the mixed utterances across the five types of blending. In addition, we tried to establish the matrix language for every utterance. Some of the utterances showed a clear matrix language (NGT or Dutch, see 2.1.4), while for others, there was no clear matrix language (indicated in Tables 3 and 4 as ø Matrix).
Looking at the rightmost columns of Tables 3 and 4, we can see the distribution of matrix types, and looking at the bottom row, we can see the distribution of blending types. The numbers in the dark blue columns show the amount (n) of a blending type, classified according to the established matrix language. Next to these dark blue columns the light blue columns show the percentage (%) of the total amount of the blending type compared to the total amount of matrix type, shown in the right (red) column. The red column right from the Table also shows the total percentage of matrix types used, not classified according to any blending type. The red row at the bottom of the Table contain information about the total amount (n) and percentage (%) of a blending type per type of blending, not classified according to any matrix type.
Table 3: Percentages of code-blended utterances in the output of child 1.
(Lexical) Insertion Alternation Independent Blending Congruent Lexicalisation Blended Blending Total % of matrix types n % n % n % n % n % n % NGT Matrix 2 33 - 0 4 67 - 0 - 0 6 19 NL Matrix 5 63 - 0 3 38 - 0 - 0 8 25 ø Matrix 8 44 1 6 1 6 2 11 6 33 18 56 Total 15 47 1 3 8 25 2 6 6 19 32 100
NGT = Nederlandse Gebarentaal, Sign Language of the Netherlands NL = Nederlands, Dutch
32/69 = 46%/100%
Table 4: Percentages of code-blended utterances in the output of child 2.
In the following, we shall illustrate the different types of code-mixing by means of some examples from utterances produced by the children. The examples are classified according to the five types of mixing presented in Tables 3 and 4.
(Lexical) Insertion produced by the children
Most of the utterances produced by both children can be classified as Lexical Insertion. Child 1 produced 32 mixed utterances, with a total of 15 utterances (47%) belonging to this type of blending. These 15 utterances can be divided according to the matrix language used. Eight of the utterances within the Lexical Insertion classification did not have a clear matrix language. An example is utterance C1.64, in example (11), in which the word ‘plus’ is produced in Dutch as well as in NGT. Since the utterance contains only one lexical element, the matrix language cannot be determined.
(11) Lexical Insertion with ø matrix language
Spoken plus plus
Signed PLUS (wrong sign) PLUS
‘plus plus plus’
Five of the utterances within the Lexical Insertion classification had Dutch as a matrix language. Examples of these are utterances C1.32 and C1.52, which can be seen in examples (12) and (13).
(12) Lexical Insertion with Dutch matrix language
Spoken lekker tis zoete peper
Signed NICE
‘Yummy, it’s sweet pepper.’ (13) Lexical Insertion with Dutch matrix language
Spoken wil je chocola hebben
Signed HAVE
‘Do you want chocolate?’
(Lexical) Insertion Alternation Independent Blending Congruent Lexicalisation Blended Blending Total % of matrix types n % n % n % n % n % n % NGT Matrix 7 64 3 27 1 9 - 0 - 0 11 16 NL Matrix 11 85 2 15 - 0 - 0 - 0 13 19 ø Matrix 12 27 3 7 1 2 - 0 29 64 45 65 Total 30 43 8 12 2 3 0 0 29 42 69 100
NGT = Nederlandse Gebarentaal, Sign Language of the Netherlands NL = Nederlands, Dutch
69/84 = 82%/100%
In both utterances, the occurring sign is semantically congruent with one spoken word an adjective in (12), a verb in (13), the whole proposition is expressed entirely in spoken words. Also, the expressed signs do not contribute additional meaning to the utterance.
Furthermore, there were two utterances with NGT as matrix language (utterances C1.38 and C1.69, presented as examples (14) and (15) below) in which the occurring word is semantically congruent with one sign and the whole proposition is expressed entirely in signs. Also, the expressed words do not contribute additional meaning to the utterance. (14) Lexical Insertion with NGT matrix language
Spoken hard stoppen
Signed HARD REINS REINS (gesture for reins of the horse) STOP
‘Pulling the reins of the horse hard to make it stop.’ (15) Lexical Insertion with NGT matrix language
Spoken die
Signed KNOW INDEX3a MAKE SOUND HONKING
‘(I)know that makes a honking sound.’
Child 2 produced 69 mixed utterances, with a total of 30 utterances (43%) that can be classified as Lexical Insertion. The utterances produced by child 2 show a similar pattern as the utterances produced by child 1, when divided according to the matrix language used (see the rightmost dark red columns in Table 3 and 4). Twelve of the utterances produced by child 2 within the Lexical Insertion classification did not have a clear matrix language. Eleven of the utterances within the Lexical Insertion classification had Dutch as a matrix language and there were seven utterances with NGT as matrix language.
Alternation produced by the children
In the 69 utterances produced by child 1, there was only one utterance (number C1.49, see
example (16)) in which the utterance was interrupted in order to sign BEFORE (‘before’),
resulting in an Alternation pattern. (16) Alternation
Spoken zelfde ook gemaakt (whispered)
Signed SAME BEFORE
‘(I) made the same before’
This utterance does not have a clear matrix language and is a type of utterance in which both the signs and spoken words are necessary to make up the full proposition. This type is defined by Van den Bogaerde & Baker (2005) as a mixed code-blend, also mentioned in 2.1.4.
As mentioned before (chapter 2) the two modalities are simultaneously available to bilingual speakers of a sign and a spoken language, and this results in the possibility to use two modalities at the same time. Therefore, it is not surprising that only one utterance with an alternation pattern is found, after all, bimodal bilinguals do not need to sequentially use their two languages. This is in line with what has been described by Donati & Branchini (2012) and Emmorey et al. (2005), see chapter 5 for a further comparison.
Interestingly, child 2 used more Alternation in her language production. In the 69 mixed utterances produced by this child, there were eight utterances (12% of all mixed utterances) following an Alternation pattern. In six of these utterances, a matrix language, Dutch or NGT, could be determined because either articulated signs or spoken words clearly did not contribute additional meaning to the utterance. The production of Alternation and the use of a particular matrix language might have been influenced by the language input directed towards the child, this is further discussed in 5.2. In the following two examples (C2.43 and C2.76), the use of NGT (17) and Dutch (18) as a matrix language can be seen.
(17)Alternation with NGT as matrix language
Spoken baby
Signed DEAR BABY
‘Dear baby.’
(18) Alternation with Dutch as matrix language
Spoken alles schoon mooi
Signed EVERYTHING CLEAN
‘Everything (is) clean, (that’s) nice.’
In example (19) (C2.51) no matrix language could be determined because both the expressed signs and spoken words did contribute additional meaning to the utterance.
(19) Alternation with ø matrix language
Spoken beker bij deze kleur zo twee lepeltjes thee bij thee kan wel
Signed INDEXTEA CAN INDEXMOTHER CAN yes
‘The cup (matches) this color like this, these two spoons for the tea with the tea, yes, you (i.e. mother) can.’
Independent Blending produced by the children
Eight utterances (25% of the total amount utterances) produced by child 1 and two utterances (3% of the total amount utterances) produced by child 2 showed an Independent Blending pattern, as can be seen in Table 2. Clear examples are utterances C1.66 and C1.67 (examples (20) and (21) below), in which NGT is the matrix language with signs adding some information to the spoken word. Both utterances involve the simultaneous production of
two independent and apparently autonomous monolingual utterances, as both bear meaning by themselves. However, in both utterances, the signs provide more detailed information, either about the specific location of the object the child is playing with (20) or about what the object that is being described looks like (21). For instance in example (20),
the sign INDEX3a specifies the location of the object, whereas the spoken word rondje does
not contain any additional information apart from the form of the object. (20) Independent Blending, NGT matrix language.
Spoken rondje
Signed ROUND-CL INDEX3a
‘That’s a round (object).’
(21) Independent Blending, NGT matrix language.
Spoken ovaal p-p-p
Signed OVAL-CL CURLED-SQUARE-CL
‘A curled oval makes a square.’
Congruent Lexicalization produced by the children
For the fourth category, Congruent Lexicalization, two examples (6% of all utterances) were found in the utterances produced by child 1, and none in the utterances produced by child 2. This low percentage is expected because the languages used, Dutch and NGT, differ in phonological, morphological and syntactic domains (Van den Bogaerde & Baker 2005). Therefore, it is not surprising that there are only very few blended strings that happen to display the same word order in both modalities (according to the specific grammars in the two languages, explained in 2.1.2), or even none at all, as is true for child 2.
Example (22) (utterance C1.29 in appendix 1) shows an utterance (a noun phrase, not a clause) in which the two languages do display the same word order. Also, this is an example of what Van den Bogaerde & Baker (2005) called code-blended full, as the full proposition is expressed in both modalities.
(22) Congruent Lexicalization & code-blended, full.
Spoken aardbeien thee
Signed STRAWBERRY TEA
‘strawberry tea.’
Blended Blending produced by the children
Six of the 69 utterances (19% of all utterances) produced by child 1 can be classified as a Blended Blending type. The utterances produced by child 2 contain even more Blended-Blending, 29 (or 42%) of all utterances are of this type. Examples (23) and (24) (respectively
utterances C1.39 and C1.42 in appendix 1) show two of these interesting utterances that are complete and meaningful only if the fragments distributed across the two modalities are put together in a blended proposition. In example (23), the utterance in Dutch zweep dan kan je
met de hand (literally: ‘whip than can you with the hand’) is semantically incomplete. The
same can be said about the NGT utterance WHIP HARD (literally: ‘hit hard’), which is
incomplete by itself. Both the spoken and signed information is necessary to form a
complete and meaningful utterance. In example (24),the sign CAT (‘cat’) is required to get a
well-formed utterance that indicates who is sleeping; thus the sentence is complete and meaningful only if the fragments distributed across the two modalities are put together in a blended utterance.
(23) Blended Blending
Spoken zweep dan kan je met de hand
Signed WHIP HARD (meaning ‘faster’)
‘You can whip a horse with your hand to ride faster.’ (24) Blended Blending
Spoken oke ja mama slaapt
Signed CAT
‘O.k. yes mother the cat sleeps.’
4.2 Quantitative results of code-blending types produced by the mothers
A brief look at the results of the mothers shows some interesting results, presented in Tables 5 and 6 (see 4.1 for further explanation on how to interpret the Tables).
Both mothers used code-blending in their utterances (77% and 87% in all utterances, respectively), and both mothers mostly used Lexical Insertion in their blended utterances. Furthermore, the mother of child 1 (M1) used a clear matrix language in Lexical Insertion blended utterances (either NGT matrix or NL matrix and some ø matrix).
Table 5: Percentages of code-blended utterances in the output of M1, mother of child 1.
(Lexical) Insertion Alternation Independent Blending Congruent Lexicalisation Blended Blending Total % of matrix types n % N % n % n % n % n % NGT Matrix 34 100 - 0 - 0 - 0 - 0 34 49 NL Matrix 10 100 - 0 - 0 - 0 - 0 10 14 ø Matrix 7 27 11 42 1 4 1 4 6 23 26 37 Total 51 73 11 16 1 1 1 1 6 9 70 100
NGT = Nederlandse Gebarentaal, Sign Language of the Netherlands NL = Nederlands, Dutch
70/91= 77%/100%
Table 6: Percentages of code-blended utterances in the output of M2, mother of child 2.
Both mothers also show a lot Alternation in their utterances. This is remarkable because both mothers are deaf and one would therefore expect that they prefer the use of NGT over the use of spoken Dutch. The mother of the second child (M2) even uses some NL matrix in her alternated utterances, which is an even more striking result because this implies the use of Dutch sentences without any support of NGT. However, both mothers are in
communication with their hearing child and therefore a blended use of NGT and Dutch is to be expected.
Another interesting result is the use of Blended Blending displayed by both mothers. As described in 2.1.2, this type of blending consists of a mixed utterance of which the constituents are distributed across the two channels. Crucially, the utterance is complete and meaningful only if the fragments distributed in the two channels are put together in a unique, blended utterance. This type of blending is to be expected in utterances of bimodal bilinguals because both constituents distributed across the two channels are necessary for a
meaningful proposition, and it seems that this requires advanced knowledge of both
languages. For people with no (complete) access to the modality of a spoken language, as is the case for both deaf mothers, it is difficult to have enough knowledge of Dutch to use Blended-Blended utterances. Therefore, it is striking that both mothers did produce this amount (9% and 10% in all utterances, respectively) of Blended Blending.
(Lexical) Insertion Alternation Independent Blending Congruent Lexicalisation Blended Blending Total % of matrix types n % n % n % n % n % n % NGT Matrix 21 84 4 16 - 0 - 0 - 0 25 34 NL Matrix 6 67 3 33 - 0 - 0 - 0 9 12 ø Matrix 22 56 8 21 1 3 1 3 7 18 39 53 Total 49 67 15 21 1 1 1 1 7 10 73 100
NGT = Nederlandse Gebarentaal, Sign Language of the Netherlands NL = Nederlands, Dutch
73/84= 87%/100%
5. Discussion
In this section, some notable results are discussed, starting with some striking results found for the blending types the children produced (section 5.1). In section 5.2, the code-blending types produced by the mothers and the children will be compared to determine whether any patterns occur in these results.
5.1 Analysis of code-blending used by the children
As can be seen in Tables 3 and 4, a considerable number of utterances produced by the
children display some type of blending, respectively 46% and 82% of all utterances. This is a striking result considering the setting of the recordings: a hearing child playing and communicating with her deaf mother. Therefore, one might expect more use of NGT. However, this result is in line with previous findings (Van den Bogaerde 2000). The high amount of blended utterances produced by the children might be influenced by the language input, the age of the participants in the present study, by their hearing status, since both children were hearing, and by their ability in both languages. Baker & Van den Bogaerde (2008) discussed results of language use of a hearing child with a similar age as the children in the present study (respectively age 5;11 and 5;9, compared to age 6;0) in conversation with his deaf mother. According to their results, this child also displayed a high percentage of code-blending (70%) in conversation with his deaf mother.
Another notable result is the frequent use of Lexical Insertion (an autonomous, complete utterance in one modality, supported by a few words or signs in the other modality) and Independent Blending (the simultaneous production of two independent, autonomous monolingual utterances, the two sentences may be slightly different in meaning). Donati & Branchini (2012) described these types of blending, along with the other types mentioned in chapter 2, summarized in Table 1. However, they do not quantify their data. Therefore, it is not possible to compare the percentages found in this study with the study conducted by Donati & Branchini (2012). Research carried out by Baker & Van den Bogaerde (2008), on the other hand, did not focus on these types of blending and therefore does not provide material to compare the results of both studies. However, a possible interpretation of the frequent use of Lexical Insertion and Independent Blending might be that it may be easier to comprehend and/or produce these types of blending simultaneously. For instance, Congruent Lexicalization (occurring in 6% of the utterances produced by child 1 and in 0% of the utterances produced by child 2) requires a chance concurrence of the structure of two languages in different modalities, displaying the same word order according to their specific grammars.
Another interesting result is constituted by the 19 percent of Blended-Blending in the utterances of Child 1 and 42 percent in the utterances of child 2. This is remarkable because it implies that at least child 2 often correctly computed non-linearized constituents
and integrated them into one global utterance, which is cognitively a difficult task. Donati & Branchini (2012) claim to have found similar Blended-Blending utterances in their study, although they do not provide any concrete numbers or percentages. However, they do argue that utterances displaying this type of blending imply that a single utterance can be distributed across two channels without linearizing some of its main constituents. In future (neuro-imaging) research, it would be interesting to look for what kind of activation patterns can be found in the brain while a Blended-Blending is uttered.
5.2 Analysis of code-blending: a comparison of mother and child
It is interesting to relate the language input provided by the mothers to the children’s language output in this study, given that previous studies (discussed in chapter 2) showed that the amount of code-blending in the language input of the children influenced the children’s output. Therefore, in 5.2.1 the amount of code-blending occurring in the input from the mothers directed to the children is quantified, and the results are related to the amount of code-blending observed in the output of the children.
Also, it is interesting to compare the matrix language of the utterances from both the mothers, as well as from the children, to determine if one of the languages is used predominantly. These results are discussed in 5.2.2.
Furthermore, in order to investigate the role of the linguistic factors influencing the code-blending occurring in the utterances, it is interesting to carry out a further analysis at the lexical level of the types of word/sign classes with which code-blending occurred. This analysis will be discussed in 5.2.3.
5.2.1 Language input from the mothers compared to language output of the children
As for the input provided by the mothers, we found similar patterns for both mothers, see Figure 1. Both mothers used a comparable amount of code-blending and blending types. Lexical Insertion was used in most code-blended utterances with a total amount of 51 uttered by mother 1 and a total of 49 uttered by mother 2. Both mothers also produced alternating utterances with a total amount of 11 for mother 1 and 15 for mother 2. Furthermore, both mothers used some Blended Blending (with a total amount of 6 uttered by mother 1 and 7 Blended-Blended utterances used by mother 2). However, they used the code-blending types Independent Blending and Congruent Lexicalisation only once.
Figure 1: Amount of code-blending occurring in the input from mother 1 (left) and 2 (right).
The results found for the children’s output show a different pattern from the results found for their mothers (see Figure 2). Both children mostly uttered code-blended utterances with a Lexical Insertion blending type. However, they also frequently uttered others types of code-blending. Especially the frequent use of Blended-Blended utterances produced by both children and the frequent use of Independent Blending uttered by child 1 (see Figure 2 left) and Alternation uttered by child 2 are interesting results.
Figure 2: Amount of code-blending occurring in the output of child 1 (left) and 2 (right).
These results do not only show a different pattern found for the utterances produced by the mothers, but they also show a difference in the output produced by the children. Where one of them produced a lot of a given code-blending type (Independent Blending produced by child 1 and Alternation produced by child 2), the other produced much less of this particular code-blending type. For instance, only one instance of (sequential) code-switching (i.e. Alternation: occurring in 3% of the utterances of child 1) was found in the total amount of utterances analysed in child 1. This result corresponds to the low 5% found in Emmorey et al.’s (2005) corpus from adult signers, which lead the authors to the conclusion that code-switching is unusual when the two modalities, sign and speech, are involved. Emmorey et al.
51 11 1 1 6
