THE ASSESSMENT OF VOCABULARY DEVELOPMENT IN YOUNG BILINGUAL CHILDREN
by Brigitta Keij
A Master’s thesis submitted in partial fulfillment of the requirements for the degree of
Master of Science
(Clinical Linguistics)
at the Joint European Erasmus Mundus Master’s Programme in Clinical Linguistics (EMCL)
UNIVERSITY OF POTSDAM
January, 2011
THE ASSESSMENT OF VOCABULARY DEVELOPMENT IN YOUNG BILINGUAL CHILDREN
Brigitta Keij
Under the supervision of Professor Gerard Bol at the University of Groningen, The Netherlands
ABSTRACT
Some young bilingual children score lower than the monolingual based norms on standardized tests for vocabulary. This is why bilingual children are often over identified as having a DLD when tested in only one of their languages. The aim of this study is to assess the vocabulary development of young bilingual children in both of their languages and to provide more insight into the internal and external factors that can possibly account for the large variation between young bilingual children found in previous research.
The outcomes of this study show that for two out of six children, their Dutch vocabulary is below the monolingual Dutch norms, but that this is certainly not the case when the vocabulary scores for both languages are taken into account. Therefore, this study shows how important it is to measure both of the languages of bilingual children. Furthermore, the pattern of vocabulary development that the bilingual children exhibit is similar to the pattern monolingual Dutch children follow. However, their developmental pattern can be influenced by the home language. The vocabulary in the home language is mostly influenced by the factor age. The Dutch vocabulary is mostly influenced by the quantity of the Dutch input the children receive. L1 typology only seems to have an influence on the pattern in which the children acquire the different grammatical word categories. The other factors that are predictors of the individual variation between the children are all related to the quality of the input the children are exposed to.
ACKNOWLEDGMENTS
I would like to thank the Piëzo foundation in Zoetermeer, especially the community centers in Palenstein, Meerzicht and Buytenwegh, and the day care center ‘t Mussennest, also in Zoetermeer, the Netherlands. Furthermore, I would like to thank Gerard Bol for his useful advice and guidance, and also thanks to Flavia Adani for being my second reader. I would especially like to thank Karijn Helsloot for providing me with the materials to conduct this study, and my EMCL colleague and dear friend Vanessa Wolz for checking my English and all her other useful comments. Last but not least, I would like to thank my parents who made it possible for me to be part of this Master programme.
TABLE OF CONTENTS
ABSTRACT ... ii
ACKNOWLEDGMENTS... iii
LIST OF TABLES ... vii
LIST OF FIGURES ... viii
1 Introduction ………...1
1.1 Vocabulary development in typically developing monolingual children …….3
1.2 Vocabulary development in monolingual children with a developmental language disorder ……….5
1.3 Vocabulary development in typically developing bilingual children ………...8
1.3.1 Moroccan-Arabic – Dutch ………...10
1.3.2 Tarifit-Berber – Dutch ………...13
1.3.3 Turkish – Dutch ………...13
1.3.4 Papiamentu – Dutch …………...15
1.4 Vocabulary development in bilingual children with a developmental language disorder ………....16
1.5 Research questions ………..18
2 Method ………20
2.1 Materials and procedure ………..23
2.1.1 Bilingual Lexicon lists ……….23
2.1.2 Language background questionnaire ………...26
2.1.3 Procedure ………..27
2.2 Data collection and participants ………..28
2.3 Data analyses ………...29
3 Results ……….30
3.1 Bilingual Lexicon lists ……….30
3.2 Language background questionnaire ………...38
4 Discussion and conclusions ………...44
REFERENCES………...52 APPENDICES………57
LIST OF TABLES
1.1 The most frequent first words in Dutch (Schlichting, 1996) ………..…...5
2.1 Semantic and grammatical word categories of the Bilingual Lexicon lists ……...26
2.2 Participants divided over the four language groups ………...28
3.1 The scores on the Bilingual Lexiconlists of the six young bilingual children …...31
3.2 The correlations between age and the productive vocabulary scores …………...33
3.3 The proportion of different grammatical word categories per child ………...34
3.4 The internal factors derived from the language background questionnaire …………39
3.5 The external factors related to the quantity of input in each language ………40
3.6 The external factors related to the quality of input in each language ………..42
LIST OF FIGURES
3.1 Age plotted against the different productive vocabulary scores ………..………33 3.2 The distribution of the grammatical word categories with increasing age …………...35 3.3 The distribution of the grammatical word categories for the ages 20-33 months ...36 3.4 The distribution of the grammatical word categories for the ages 35-47 months ...36 3.5 The distribution of the grammatical word categories per language group ……...37
Chapter 1 Introduction
For most children language development comes naturally, but for children with a developmental language disorder (DLD) this is not the case. The early recognition of DLDs is very important in order for early intervention to take place. This is not only important for monolingual children, but also for bilingual children, from immigrant families, for example, who speak a different language at home than the language of the community. To make the early recognition of DLDs possible, a test instrument that is suitable for testing young bilingual children needs to be developed. In the Netherlands this kind of instrument did not yet exist until the Bilingual Lexicon lists were developed in 2006 (Schlichting, 2006).
In the current situation in the education system in the Netherlands, there is a lot of concern regarding the language development of bilingual children from immigrant backgrounds. This is mostly in relation to their language ability in Dutch, but bilingual children that have problems in the acquisition of Dutch might also have problems in their mother tongue. A delay in language development can be an indication of several problems.
First of all, a language delay can be an indication of a specific language impairment, when other aspects of childhood development show no or very little problems. In addition, a
language delay can be caused by a hearing impairment or a delay in a broader cognitive development than language development alone, namely a general developmental delay. The early recognition of a language delay is a prerequisite for early treatment, and therefore very important. The language development of monolingual Dutch children is currently screened early on in development by pediatricians in child health care centers. This is done with screening instruments and after determining a language delay, the decision is made either for further consultation or reference to an auditory centre or a speech therapist. This way the treatment can start as early as possible, typically around the age of three.
In children that have another language than Dutch as their mother tongue, or even two different mother tongues, early recognition of DLDs or language screening in general is more difficult. The mother tongues of these children must be taken into account when assessing their language development. The pediatrician or speech therapist however, usually neither has the working knowledge nor the instruments to include these languages. This means that language delays and DLDs in bilingual Dutch children may be discovered later than in monolingual Dutch children. As a consequence, referral or treatment may occur later or not at all. It can also be the case that bilingual children with a language delay in Dutch receive unnecessary treatment for a DLD whilst their development in the mother tongue is unimpaired. This is why a screening instrument for this group of children was highly necessary.
Based on the Lexilist Dutch (Schlichting & Lutje Spelberg, 2002), four Bilingual Lexicon lists were developed for the following languages: Moroccan-Arabic, Tarifit-Berber, Turkish and Papiamentu (Schlichting, 2006). Each of these lists, except the Dutch- Papiamentu list, have been standardized and validated for certain age ranges. For all of the bilingual groups in the standardization research, large standard deviations have been found
within the monthly age groups and language groups which means there is a lot of variation between children (Van Agt, Gonc-Erbasi, Elhajoui, Schlichting, 2004; Boerma, 2005).
1.1 Vocabulary development in typically developing monolingual children
Most of what is known about first language development is based on large observational or experimental research groups, despite the fact that small observational or experimental research, such as case studies, can often shed more light on the individual differences in language development between children. Many theories about language development are based on the larger corpus of data collected through the years. These data show a gradual process of language acquisition which can be divided into several periods and phases. These periods and phases are primarily based on certain features that are evident in the language use of children, the so-called ‘language internal features’. It is possible to link chronological ages to these periods and phases, although there is a lot of variation between the ages at which children reach or complete certain periods and phases.
According to Schaerlaekens (2000) there are four periods and phases that can be distinguished: the prelinguistic period (age 0-1;0) in which the child makes sounds, but communicates without using meaningful words, the early linguistic period (age 1;0-2;6) in which the child uses meaningful words and combines words into meaningful phrases, the differentiation phase (2;6-5;0) in which the child makes more complete and correct sentences and the completion phase (5;0-9;0) in which the metalinguistics skills are developed.
The early linguistic period starts with the transition from babbling into the production of the first meaningful words. In this period two foundations of language are built: the active vocabulary and the basis for syntax. The development of these two foundations can be divided into two phases: the one word phase and the two word or multiple word phase. The
first word can appear between 8 and 18 months of age, but by 14 months approximately 90%
of the children have produced their first word (Fenson, Dale, Reznick, Thal, Bates, Hartung, Pethick and Reilly, 1993; Schlichting, 1996). Dutch speaking children produce approximately 50 words around the age of 1;9 (Schlichting, 1996). After reaching the 50 word threshold, a so-called vocabulary spurt might take place usually before the age of 2;0, but this spurt is not evident in all children. Between the age of 1;6 and 2;0, 95% of Dutch children start using two word phrases (Schlichting, 1996).
The active vocabulary in this phase mainly contains content words, mostly nouns, and contains almost no function words, such as determiners, prepositions, pronouns and conjunctions. Schlichting (1996) researched the most frequent early words in Dutch. Table 1.1 presents the words that were produced by at least one third of the 37 children in her research as part of their first 50 words. The numbers between the brackets represent how many words from that category were included in the research. The number in front of the word shows the number of children that produced the word and in italics is the English translation of the Dutch words.
These results show that the majority of the words in the early vocabulary are nouns.
This result confirms the assumption that a so-called noun bias might exist in the first language acquisition of Dutch, which means nouns are acquired earlier than other word categories. Some researchers assume that all children have the conceptual predisposition to words that are linked to objects which implies that the noun bias is universal for all languages. However, other research has shown that the noun bias is not universal, for example children acquiring Korean and Mandarin-Chinese do not conform to this pattern (Gopnik & Choi, 1995; Tardif, Gelman & Xu, 1999). Whether this is the case for children acquiring the languages researched in this study will be discussed in paragraph 1.3.
Table 1.1. The most frequent first words in Dutch (Schlichting, 1996).
Nouns (213) Social expressions (45)
Verbs (49) Other (58)
35 auto car 26 bah bah 20 eten eat 21 die that
32 mama mommy 25 dag hi/ bye 17 zitten sit 21 uit out/ off 32 papa daddy 24 nee no 15 slapen sleep 20 op on/ over 30 poes kitty/ cat 22 au ouch 14 poepen poop 13 buiten outside 27 opa grandpa 21 ja yes 13 drinken drink 13 open open 23 koekje cookie 15 hap bite
23 oma grandma 13 boem bang 22 pop doll 13 kiekeboe
peekaboo 21 bal ball
21 paard horse 18 hond dog 17 jas coat 17 klok clock 16 eend duck 14 appel apple 14 boek book 13 aap monkey 13 kaas cheese
1.2 Vocabulary development in monolingual children with a developmental language disorder
Developmental language disorders (DLDs) include language disorders caused by hearing impairments, language disorders that occur in co-morbidity with other disorders such as Down’s syndrome and autism spectrum disorders, as well as specific language impairment (SLI). The main focus in this paragraph will be on the last type of DLD, namely SLI. SLI is a neurodevelopmental disorder that affects approximately 7% of the population (Leonard, 1998). SLI is usually diagnosed through a combination of inclusion and exclusion criteria.
Children with SLI show language abilities below what is expected for their age but do not have any hearing impairments, a below average IQ, an autism spectrum disorder, motor problems or acquired neurological damage (Leonard, 1998). In other words, SLI is a language impairment that is not caused by deficits in other areas of cognition.
Children who are later diagnosed with SLI were often found to be delayed in the onset of their first words and first word combinations. This means that children with SLI often start out as late talkers, but although an initial delay for most late talkers is later on resolved, this is not the case for children with SLI. At two years of age, 7% of these children who are otherwise developing typically have fewer than 30 words in their expressive vocabulary, which is between a sixth and a tenth of what they should have. Children who have such small vocabularies, which are not attributable to poor hearing, low intelligence, an autism spectrum disorder or brain damage, have a lexical delay or disorder. Half of these children also have a restricted receptive vocabulary and it is this group of children who have the biggest chance of ending up as language impaired. The children that are only delayed in their productive vocabulary and have no limited receptive vocabulary are usually able to catch up, but but many of these children have problems with reading, writing and spelling as adults (Locke, 1997).
Research suggests that many children with a small receptive and expressive vocabulary may be four to six years old before they acquire as many words as used by a typically developing two-year-old (Locke, 1997). For the early detection of developmental language disorders, such as SLI, it is important to assess the development of both the productive and receptive vocabulary of children. However, it is very hard to measure the receptive vocabulary of children through parental report, since research has shown that parents have difficulty keeping track of how many words their child understands after the age of 1;4 (Bates, Marchman, Thal, Fenson, Dale, Reznick, Reilly & Hartung, 1994).
Although children with SLI form a heterogeneous group, there are some profiles of impairments that reoccur in different children concerning vocabulary development. Some children do not have any language production at an age where this is expected. This is known as mutism. Other children do make sounds, but do not produce any meaningful words at ages
where meaningful production is expected. When children are considerably late in producing their first meaningful words, for example only at the age of three, this is called a delayed language onset. Moreover, there are children who experience word finding problems later on in their linguistic development (Schaerlaekens & Goorhuis-Brouwer, 2000). Early detection of a delayed vocabulary development is important for early intervention to take place, since a limited vocabulary may lead to other problems in linguistic development.
Van Santen (2010) compared the productive vocabulary development of children with SLI to typically developing children and found that globally speaking, their development is similar: the majority of the early vocabulary exists of nouns and as the mean length of utterance (MLU) gets longer, the proportion of nouns reduces. Furthermore, for both groups, the production of verbs is relatively low early on in development but increases later on.
However, there is a significant difference in the quantity of nouns and function words between the children with SLI and the typically developing children. Typically developing children produce significantly more noun types than children with SLI. Additionally, the decline of the proportion of nouns in the vocabulary is more drastic in typically developing children than in children with SLI. A similar trend is visible when looking at function words:
the use of function words increases more rapidly in typically developing children than in children with SLI. In sum, children with SLI exhibit the same vocabulary development as typically developing children, except for the fact that the changes in the proportion of nouns and function words is less distinct for the children with SLI.
1.3 Vocabulary development in typically developing bilingual children
In the current study, bilingual children are assessed on their vocabulary development in both of their languages. Bilingualism in children can be classified in different ways. When children acquire two languages before the age of 3;0, usually from birth, they are called simultaneous bilinguals (2L1). When children start learning the second language after the age of 3;0, after the first language has already been partly established, they are called sequential bilinguals (L2) (Genesee, Paradis, & Crago, 2004).
These terms can be linked to a classification according to language competence.
Simultaneous bilinguals are assumed to become balanced bilinguals, which means their competence is roughly similar in both languages, whereas sequential bilinguals are assumed to become dominant bilinguals, which means one of the two languages is dominant.
However, research has shown that young bilingual children, simultaneous as well as sequential, are usually more proficient in one language. The dominant language tends to be the language they have received the most amount of input from (Genesee et al., 2004).
Research by Pearson, Fernandez, Lewedge & Oller (1997) among Spanish-English bilingual infants has shown that at least 20 % of exposure time devoted to one language is needed to develop an active lexicon, while a balanced form of bilingual vocabulary development requires 40-60 % exposure to each language.
However, the amount of input is not the only factor that influences the language development of a bilingual child. Research has shown that factors including language aptitude, first language (L1) typology, age of acquisition, parent education, and quality of input can also influence the rate and success in bilingual language acquisition (see Paradis, 2007, for an overview). Paradis (2010) conducted another study to find out which combination of these factors predicts the development of vocabulary size and verb morphology in early English second language (L2) acquisition, whether certain factors
change in early or later stages of acquisition and whether the same factors predict lexical and morphological acquisition.
The children in her study were divided into two groups: early stage (< 18 months exposure to English) and later stage (> 18 months exposure to English). A multiple regression analysis was conducted within each group for vocabulary and morphology scores separately.
The results showed that the child's internal factors are strong predictors in the early stage group, both for vocabulary and verb morphology, whereas the child's external factors, such as the richness of the English environment, are strong predictors for both aspects in the later stage group. Maternal education was significantly and positively correlated with the mother’s fluency in English, but maternal education was not a significant predictor of vocabulary or verb morphology at either stage. Research by Boerma (2005) has shown that within a Moroccan-Arabic – Dutch language group, a significant predictor of vocabulary development in Dutch was whether the parents or caretakers of the child had taken Dutch lessons. Taking Dutch lessons was positively correlated with the child’s vocabulary development in Dutch, for both the father and the mother.
The distinctions between different kinds of bilingual development mentioned above can also be linked to a more cognitive classification of bilinguals into ‘compound’ and
‘coordinate bilinguals’. This classification is especially of interest when looking at the vocabulary development of bilingual children because it addresses the question of whether bilinguals have two different representations for one word or concept, or if they have a shared representation for one concept. This question can be answered by investigating whether a bilingual child acquires one language system or two separate language systems. According to Clark’s principle of contrast (1987, in Pearson, Fernández & Oller, 1995) children assume that each word form must have a different meaning. This means they use the word form for a
referent of the language in which they first learned the concept. This principle would apply until children have acquired 150 words.
However, there is also evidence that young bilingual children already use cross- linguistic equivalents before this 150 word threshold. Besides, the fact that they use a word in one language, but not in the other could also point to the fact that the word is not present in the input of one language. Pearson et al. (1995) found that in about 30% of all concepts known by the bilingual child, there was a translation equivalent, which means the word was produced in both languages. This was the case for children with a vocabulary of 2-12 words as well as children with a vocabulary up to 500 words.
Appel & Vermeer (2000) state that most studies show that bilingual children lag somewhat behind in each language compared to monolingual children that acquire the same language. This is also the case for the vocabulary development: the vocabulary in each language is smaller, but when all known words for both languages are added, the total vocabulary is usually bigger than that of monolingual children. Consequently, different researchers have shown that typically developing bilingual children only show a delay in language development when they are assessed in only one of their languages. Pearson, Fernández, & Oller (1993) conclude that the pattern of lexical development in the two languages of bilingual children is very similar to that of the development in monolingual children. They are not slower in the development of their productive vocabulary than monolingual children. Therefore, it is necessary to take into account both languages in assessing the vocabulary development of bilingual children.
1.3.1 Moroccan-Arabic - Dutch
Boerma (2005) conducted a large study using the Bilingual Lexicon lists in order to standardize a Moroccan-Arabic – Dutch lexicon list. The parents or caretakers of 232 young
bilingual children who received input from Moroccan-Arabic at home participated in the study. The children were between 18 and 34 months of age, with an average age of 24 months. The amount of words in the vocabulary increased with the age of the children, which is expected. However, the correlation was not very strong, due to the large variation between the children. Only part of this variation could be explained by the factor age, and therefore there must have been other factors involved, which will be discussed later. There were no differences found between girls and boys in the Moroccan-Arabic – Dutch group. These differences were found in the Dutch control group, with girls scoring higher than boys (Schlichting & lutje Spelberg, 2002). In comparison to this Dutch control group, the bilingual Moroccan-Arabic – Dutch group scored lower, both when the total amount of words and when the total amount of concepts were taken into account. There was a visible trend towards becoming more bilingual with age, as the younger children often produced a word in only one language and the older children used the words from both languages for one concept.
The external factors that were investigated by Boerma (2005) as to their influence on the language development of young bilingual children are: Dutch lessons taken by the parents, the languages spoken by each parent, the dominant language spoken with the child, and whether the child has older siblings. The results showed that when parents have taken Dutch lessons, this has a positive influence on the Dutch vocabulary score. On the other hand, there was a relation between the languages spoken by the father and the vocabulary score in Moroccan-Arabic. Children with fathers that spoke Moroccan-Arabic and Berber, scored higher in Moroccan-Arabic than children with fathers that spoke Moroccan-Arabic and Dutch. There was no relation found between the dominant language spoken with the child and the vocabulary score in Moroccan-Arabic, but there was a relation with the vocabulary score in Dutch. Children who were mainly spoken to in Moroccan-Arabic scored lower in
Dutch than children who were also or mostly spoken to in Dutch. Whether a child had older siblings was not of influence on the vocabulary score in either language.
In accordance with the literature, the proportion of nouns, onomatopoeia and routine words such as sayings and exclamations, decreased with age. The proportion of verbs, closed class words and adjectives increased with age for both the Moroccan-Arabic and the Dutch vocabularies. For the early vocabulary, the number of nouns was bigger than the number of verbs in both Moroccan-Arabic and Dutch. However, in Moroccan-Arabic the difference between the proportion of nouns and verbs was not very large, and this trend reversed once the vocabulary grew beyond 125 words with the number of verbs exceeding the number of nouns in Moroccan-Arabic. This trend is not visible for the Dutch vocabulary, where the percentage of nouns continues to exceed the percentage of verbs later on in development.
Boerma (2005) explains this by referring to the characteristics of Moroccan-Arabic.
Moroccan-Arabic has a VSO word order, which means that the verb has a more prominent position in each utterance than in Dutch, which is a verb-second (V2) language. Futhermore, Moroccan-Arabic is a pro-drop language, which means that the verb carries more linguistic information than in a language that has obligatory use of pronouns, such as Dutch. Moreover, the use of imperatives differs in Moroccan-Arabic and Dutch. In Dutch, parents use a short phrase which includes a noun such as, ‘Go to your bed!’ while in Moroccan-Arabic parents simply use the verb, ‘Go!’ (E-rramdani, 2003). All of these characteristics make verbs more prominent in the Moroccan-Arabic input that children receive, and these results can be seen as counter evidence of the noun bias being a universal phenomenon, which for Moroccan- Arabic does not seem to last very long in language development.
1.3.2 Tarifit-Berber – Dutch
Van Agt et al. (2004) conducted a large study using the Bilingual Lexicon lists to compare the word scores of bilingual Tarifit-Berber - Dutch speaking children with the norms of monlingual Dutch children, and to develop norms for this group of bilingual children between the ages of 24-30 months. The distribution of word scores resembled the distribution of word scores in monolingual Dutch speaking children, but the percentage of children with very low scores was high.
An earlier study by E-rramdani (2003) investigated the language development of both Tarifit-Berber and Dutch in bilingual children between the ages of 4-6 years. The study encompassed, among other domains, the receptive and productive vocabulary of the children.
The results showed that the best scores were reached on the receptive vocabulary task for Dutch, while the lowest scores were achieved in productive vocabulary in Dutch. With respect to language dominance, the 4-year-old children turned out to be balanced bilinguals, whereas the 5- and 6-year-olds were dominant in Dutch. These results could also be linked to the external factor of language use: the language use patterns of the children showed that 85% of them spoke Tarifit-Berber with their parents, while the two languages were used equally with siblings. The age groups differed in their language use with friends: 66% of the 4-year-olds spoke Tarifit-Berber with their friends, while the 5- and 6-year-olds only spoke Dutch with their friends.
1.3.3 Turkish – Dutch
In the large study by van Agt et al. (2004) using the Bilingual Lexicon lists, young bilingual Turkish – Dutch children were also included in order to develop norms for the Turkish – Dutch lexicon list for the ages of 24-30 months. The distribution of the word scores for this
group also resembled the distribution of the word scores in monolingual Dutch children, but again there was a high percentage of children with very low results.
In a recent study by Leseman (2010) on the bilingual vocabulary development of Turkish preschoolers in the Netherlands, the development of the receptive and productive Turkish and Dutch vocabulary of 31 Turkish – Dutch bilingual children aged 3-4 years was assessed and compared with the Dutch vocabulary development of 77 monolingual Dutch children of the same age. Within the bilingual and monolingual group, the children were matched on socio-economic status (SES). The bilingual children’s vocabulary development in Turkish appeared to be similar to the monolingual low SES children’s vocabulary development in Dutch. However, the bilingual children’s vocabulary development in Dutch was further behind. The vocabulary of the bilingual children in Dutch grew significantly in their fourth year of life because of the increased influence of preschools and the introduction into school at age four. This clear increase occurring in the Dutch vocabulary did not occur in the Turkish vocabulary of the bilingual 4-year-olds.
The fact that the bilingual preschool children were dominant in Turkish is due to common trend of Turkish families providing almost homogeneous Turkish language environments to their preschool-aged children. Leseman also found that there was a strong relation between the general cognitive ability of the bilingual children at age 3;2 and their vocabulary in Turkish later on, indicating that language learning is also partly internally driven. However, general cognitive ability as an internal factor correlated less strongly with their Dutch vocabulary scores. These results correspond to the finding of Paradis (2010) that the child's internal factors are strong predictors of vocabulary development in the early stages of language acquisition which in this case was almost exclusively the acquisition of Turkish.
The child's external factors, such as the richness of the language environment, are strong predictors in the later stages, when the acquisition of Dutch started.
1.3.4 Papiamentu – Dutch
The standardization of the Papiamentu – Dutch lexicon list has not yet occurred, and therefore there is no literature available on studies conducted using the Bilingual Lexicon lists for this group of young bilingual children. However, Narain (1995) conducted a large study on the acquisition of Papiamentu and Dutch in Curacao and in the Netherlands with children between the ages of 4-6 years. This study encompassed, among other linguistic tests, tests for receptive and productive vocabulary. These tests were administered three times: once shortly before the children first went to school, once after one year of schooling and the last time after 2 years of schooling. For the bilingual children living in the Netherlands, the results showed that the productive vocabulary is relatively low for both languages during the whole period of testing. The development of the productive vocabulary in Papiamentu and Dutch clearly shows that in the first year of schooling, the two languages have a similar development, whereas in the second year of schooling, Dutch takes the lead. At the end of the two years the children scored higher on Dutch than on Papiamentu. This also seems to be caused by the fact that the vocabulary development in Papiamentu stagnates in the second year of schooling. The receptive vocabulary develops in a more parallel fashion for the two languages.
Before these bilingual children go to school, the factors; home language, language choice and cultural orientation appear to have a positive relation with the language development in Papiamentu. The amount of time the parents of the child have lived in the Netherlands and preschool experience appear to have a negative relation with the language development in Papiamentu. The longer the parents are in the Netherlands, the less they speak Papiamentu with their child. At the end of the first two years of schooling, the factors;
home language, cultural orientation, and duration of stay of the parents, seem to be the most
important predictors of the development of Papiamentu. On the other hand, the factors;
duration of stay and cultural orientation of the parents, and the language contact and choice of the child are positively related to the Dutch language development. The factors; SES, language contact and choice of the child, and extra Dutch lessons in school, seem to be the strongest predictors for the language development in Dutch.
1.4 Vocabulary development in bilingual children with a developmental language disorder
When a, for example, bilingual Turkish – Dutch speaking child has a language delay in Dutch compared to other Dutch speaking children of the same age, this could point to two things:
either the child is only delayed in Dutch, but not in Turkish, which means the child probably does not have a developmental language disorder (DLD), or the child is delayed in both languages, which could mean the child has a DLD. Therefore, in order to diagnose a bilingual child with a DLD, the child should always be diagnosed in both languages. Theoretically, DLD’s are just as likely to occur in bilingual children as in monolingual children. However, bilingual children are often overrepresented in special education or language therapy services (Cummins, 2000; Donovan & Cross, 2002; Klinger & Artiles, 2003). Over identification of SLI in bilinguals could be caused by the overlap in linguistic characteristics between typically developing (TD) bilingual children and monolingual children with SLI, and between TD bilingual children and bilingual children with SLI. This overlap is most obvious in the early stages, the first 2 years, of bilingual language acquisition (Paradis, 2010).
An important question is how TD bilingual children perform on standardized tests developed for monolinguals and whether their performance would fall below the normal range of TD monolingual children, resulting in possible misidentification. Thordardottir,
Rothenberg, Rivard & Naves (2006) found that French – English bilinguals aged 2;6 scored lower than the monolingual based norms on standardized tests for vocabulary and syntactic development in both languages. Patterson & Pearson (2004) report on similar findings for vocabulary development in Spanish – English toddlers, but in terms of total conceptual vocabulary, bilinguals do not lag behind monolinguals. However, bilinguals might lag behind monolinguals longer for vocabulary-based measures in one of their languages (Oller, Pearson, & Cobo-Lewis, 2007). Oller et al. (2007) showed that Spanish-English bilingual children’s scores for tests of receptive and productive vocabulary fell below the normal range.
Rescorla & Achenbach (2002) studied the vocabulary development of 278 children between 18 and 35 months old. They compared the scores of monolingual English speaking children to those of bilingual, also English speaking, children from different ethnic backgrounds. The degree of language delay, for which they used a cut-off of less than 50 words or the absence of word combinations, was higher for the bilingual groups (29 and 24%) than for the monolingual group (4%). Patterson (2004) comments on the results of Rescorla & Achenbach by saying that, although the authors administered the Language Development Survey (LDS) by instructing the parents of bilingual children to mark a word on the form if the child uses the word in English or if the child uses an equivalent word in the other language, this method could still have resulted in an underestimation of these children’s vocabulary size, because the parents had to read the list of English words, translate the words to the other language and then decide if they had heard their child produce the word in either language.
Studies by Conti-Ramsden, Crutchley & Botting (1997) and Crutchley (1999) compared the English language development of bilingual children with SLI and monolingual children with SLI of 8 years old. Overall, the bilingual children scored lower than their
monolingual peers, both with SLI, on standardized language assessments in the domains of vocabulary and grammar. Steenge (2006) conducted a study that compared 6- and 7-year-old typically developing bilingual Turkish – Dutch children (BI-TD) and bilingual Turkish – Dutch children with SLI (BI-SLI) on, amongst other linguistic tasks, their productive vocabulary. The results show a significant difference between the productive vocabulary score of the BI-TD group and the BI-SLI group in Turkish, with the BI-TD group scoring almost twice as high as the BI-SLI group. To conclude, it seems like there is an extra disadvantage for bilingual children with SLI compared to monolingual children with SLI and typically developing bilingual children in vocabulary development.
1.5 Research questions
The aim of the current study is to assess the vocabulary development of young bilingual children and to provide more insight into the internal and external factors that can possibly account for the large variation between young bilingual children found in previous research.
This aim can be rephrased into several research questions: (i) How do the scores of the bilingual children on Dutch compare to the existing monolingual Dutch norms? (ii) Do bilingual children have below normal scores for Dutch and for both languages together? (iii) Which internal and external factors are possibly responsible for the large individual variation found between young bilingual children?
The expectations are that (i) the bilingual children might lag somewhat behind if only the Dutch language is taken into account compared to monolingual children acquiring Dutch, (ii) that based on the literature (Pearson, Fernández, & Oller, 1993; Patterson & Pearson, 2004), the score for one of the child’s languages could be below the monolingual norms, but the score for both languages together should not be below the bilingual norms in the case of
typical development and (iii) that based on the literature (Paradis, 2007), the existing variation can be explained by internal factors such as L1 typology and age of onset of acquisition, or by external factors such as parental education, and the quantity and quality of input of each language.
Chapter 2 Method
To answer the research questions Bilingual Lexiconlists, which are adaptations of the Lexilist Dutch (Schlichting & lutje Spelberg, 2002; 2004), are used. The Dutch list is based on the Language Development Survey (LDS) by Rescorla (1989). Rescorla developed the LDS to be able to detect a language delay in children of approximately two years of age. According to Rescorla, there were no user friendly instruments for children of this age. Therefore, the author decided to develop an instrument which works with parental report. The instrument consists of a list of words and a few questions on the production of word combinations. The words that are on the list were chosen following the research on vocabulary development in English-speaking children. The words are divided over different semantic categories, comprising words that are highly frequent in the vocabulary of young children, but also those of lower frequency. This was done because a ceiling could otherwise be easily reached by older children. The LDS has been reviewed and revised several times and now contains 309 words falling under fourteen semantic categories. The list works as follows: parents read the words on the list and cross off the words that their child produces. In addition, they are asked to write down three examples of the longest word combinations the child produces.
Schlichting & lutje Spelberg (2002) decided to develop a Dutch language screening instrument based on the LDS by Rescorla, namely the Lexilist Dutch. This is not a literal translation of the LDS, but a new original list. The list includes words and short sentences that young Dutch children produce, and is used to assess the language development of children between the ages of 15 and 27 months of age. A delay in language development can point to different developmental language disorders with different underlying causes. A language delay can be an indication of a specific language impairment, when other aspects of childhood development are developing typically, or a language delay can be caused by problems in other domains than just language, such as hearing problems or other syndromes.
The initial version of the Lexilist Dutch (Schlichting, van Eldik, lutje Spelberg, van der Meulen, & van der Meulen, 1995) incorporated words and sentences that had been selected following research on the first fifty words and the first short sentences produced by Dutch speaking children (Schlichting, 1996) and the additions of parents that participated in a pilot study. However, the data from the pilot study were not sufficient for standardization and a ceiling effect was measured for children of 27 months.
This is why a new study was started that included more difficult words in order to create an improved version of the Lexilist Dutch (Schlichting & lutje Spelberg, 2002). From a list of a thousand basic words (Schlichting, 2001) 179 words were added to the list. This version of the list now contained 488 words and 11 short sentences. The list was presented to parents of children in the ages of 15, 17, 19, 23, 25 and 27 months. In total, the data of 809 children were collected. Words that were crossed off by less than 5% of the parents and all proper names were taken of the list. After these fourteen items had been removed, 485 items remained. This list was still too long and not very practical given that the list had to be read and filled out by parents. Therefore, it was decided to create two lists of 263 words and 11
short sentences. There is a small overlap between the two lists and the short sentences are completely identical on both lists.
Consequently, there is now an A and a B version of the Lexilist Dutch. The lists are comparable, but contain different words and therefore different standardization tables have to be used. In all age groups it appeared to be the case that girls scored higher than boys, therefore there are also different standardization tables for girls and boys. The Lexilist Dutch uses the parental report method. Parents are asked to tick the box in front of the words their child produces by them self, in other words, not words that are repeated or imitated. It concerns the productive vocabulary of the child, which is different from the receptive vocabulary. Research has shown that the productive vocabulary, measured with instruments like the Lexilist Dutch, correlates with the language development in general to a great degree (Schlichting & lutje Spelberg, 2002). Therefore, the productive vocabulary of the child can be seen as a measure of the language development of the child in general (Burden, Stott, Forge
& Goodyer, 1996; Berglund & Eriksson, 2000).
Schlichting & lutje Spelberg also investigated the reliability and the validity of the Lexilist Dutch. For the different age groups (15, 17, 19, 21, 23, 25 and 27 months) the correlation between the two lists (list A and list B) was calculated. The correlations appeared to be very high, namely between .97 and 1.00, which means the reliability of the lists is high.
The validity of the Lexilist Dutch was only investigated with the initial version of the list from 1995. The list was presented to 91 parents of 21-month-old children together with the Reynell Test for Language Comprehension and the Test for the Development of Words and the Test for the Development of Sentences, both from the Schlichting Test for Language Production (Schlichting, 2002). The correlations between the different tests were .42, .72 and .64 respectively. For seventeen of these children, the predictive validity of the Lexilist Dutch was investigated by presenting the list to the parents a second time six months later. The
correlations this time were respectively .64, .82 and .80. This means that both the validity and the predictive validity of the Lexilist Dutch are quite high.
In the current study the bilingual adaptations of the Lexilist Dutch are used. The Bilingual Lexicon lists are developed in four languages, namely Turkish, Moroccan-Arabic, Tarifit-Berber and Papiamentu. Three of these four lists are standardized for particular age ranges (Van Agt et al., 2004; Boerma, 2005). The Papiamentu – Dutch list is not yet standardized. The lexicon lists are filled in by the parents or caretakers of the children.
Parents were also asked to fill in a language background questionnaire to provide more information about the child’s language background, and shed light on the internal and external factors that could possibly account for the differences in vocabulary development between the children. First, a description of the material and the procedure is given, and then descriptions of the data collection and the participants follow.
2.1 Materials and procedure
The materials that are used in this study are Bilingual Lexicon lists in four languages, namely Turkish, Moroccan-Arabic, Tarifit-Berber, and Papiamentu, and a language background questionnaire in Dutch. First, the Bilingual Lexicon lists are described, followed by a description of the language background questionnaire and of the procedure.
2.1.1 Bilingual Lexicon lists
There appeared to be a great need for a language screening instrument for young bilingual children in the Netherlands (Hoogsteder, Blumenthal & Yumusak, 1999). These children often hear both Dutch and their mother tongue in their environment. If their vocabulary is assessed with the Lexilist Dutch, then actually only a part of their entire vocabulary would be assessed, namely only the Dutch part. Based on the standardization tables, these children
could be assessed as having a language delay, although in most cases they have a normal language development, given that their vocabulary is divided over two languages. Research has shown that when a bilingual child’s knowledge in both languages is taken into account, their performance is very similar to that of monolingual peers from similar backgrounds (Pearson et al., 1993).
Therefore, Schlichting started a project to adapt the Lexilist Dutch to a number of frequently spoken minority languages in the Netherlands. The following languages were chosen: Turkish, Moroccan-Arabic, Tarifit-Berber, and Papiamentu. The starting point for the development of the Bilingual Lexicon lists was the Lexilist Dutch. First, the words on this list were translated into the four languages. The researchers encountered a problem in the translation for the Moroccan-Arabic and Tarifit-Berber lists because these languages only know a spoken form and not an official written form. Therefore, it was decided to use a phonetic transcription system in the Latin script. The pilot studies that were done afterwards showed that the Moroccan participants had no problems reading this script.
The translated lists have been tested on native-speakers in Morocco, Turkey, and the Antilles. Following this, the lists have been adjusted and tested in the Netherlands. Based on the frequency of the words and short sentences that were crossed off by the Dutch parents and the words that were added by these parents, the words and short sentences for the new version of the lists were selected. Then, for each language, language specialists added more words to this version of the lists to avoid a ceiling effect appearing in older children. The extended version of the lists has been tested again in the Netherlands and the words that were not crossed off by any parent were removed. Additionally, per semantic category words were taken off the lists alternately, which means every other word was randomly removed. By doing this the final versions of the lists contain both high and low frequency words, which prevents a ceiling effect from appearing.
The final version of the Moroccan-Arabic – Dutch Lexicon list consists of 253 Moroccan-Arabic items and 247 Dutch items, the Tarifit-Berber – Dutch Lexicon list consists of 311 Tarifit-Berber items and 287 Dutch items, the Turkish – Dutch Lexicon list consists of 267 Turkish items and 275 Dutch items, and the Papiamentu – Dutch Lexicon list consists of 262 Papiamentu items and 275 Dutch items. All the items are divided over 15 semantic categories and one category of short sentences which contains multiple-word-utterances and morphologically complex words. The Turkish – Dutch list has an extra category with determiners. Table 2.1 shows that the semantic categories are linked to grammatical categories, for example the semantic category ‘doing things’ is linked to the grammatical category of verbs. The final versions of the lists can be found in Appendix I.
The standardization research of the Tarifit-Berber and the Turkish lists was conducted by Van Agt et al. (2004) at the Erasmus University Rotterdam and the research for the Moroccan-Arabic list was conducted by Boerma (2005) at a foundation called Stichting Studio Taalwetenschap. For each language, 320 participants were needed for statistical reasons. These 320 participants were divided over four age groups: 21, 23, 25, and 27 months old. For each age group, 80 participants were included. The Papiamentu list was not standardized, because there were not enough participants. The Moroccan-Arabic – Dutch lexicon list is now standardized for the ages 20-28 months, the Tarifit-Berber – Dutch list for the ages 20-30 months and the Turkish – Dutch list for the ages 24-31 months.
Table 2.1. Semantic and grammatical word categories of the Bilingual Lexicon lists Semantic category Word category Example in Dutch English translation
Play Onomatopoeia &
routine
woef woof
Many things Nouns bal ball
How is it? Adjectives vies dirty
Clothes Nouns jas coat
Outside Nouns bos forest
Animals Nouns hond dog
Doing things Verbs slapen to sleep
At home Nouns raam window
Eating and drinking Nouns ijsje ice cream
Small words Closed class words (function words, pronouns, adverbs)
ik I
Vehicles Nouns auto car
Place and time Closed class words (prepositions and adverbs)
morgen tomorrow
Body parts Nouns oog eye
People Nouns meisje girl
Sayings and exclamations
Onomatopoeia &
routine
hoera hurray
Short sentences wat is dat what is that
2.1.2 Language background questionnaire
Besides filling in the Bilingual Lexicon list, the parent or caretaker is also asked to fill in a language background questionnaire (Appendix II). Studies on child bilingualism have shown large individual differences between children acquiring two languages. Research has shown that factors including language aptitude, first language (L1) typology, age of acquisition, parent education, and quantity and quality of input can influence the rate and success in bilingual language acquisition (see Paradis, 2007 for an overview). In the current study the parents are asked to fill in a language background questionnaire from which the internal and external factors can be derived for later analyses. The language background questionnaire contains six main questions with several sub-questions.
The first group of questions is about the details of the parent completing the form:
name, gender, and the date of when the form was filled in. The second group of questions
concerns the details of the child: name, the number of siblings the child has, the age of acquisition of the Dutch language, and medical questions about whether the child was born premature and whether the child has hearing problems. The third and fourth set of questions is about the parents' language background, for the father and mother separately: nationality, mother tongue, educational level, and whether the parent has taken Dutch lessons. The fifth group of questions concerns the quantity of the input the child receives in each language: the language the parents speak to each other, the language the siblings speak to each other, but also about the language comprehension of the child and which language is most valued by the parents. For the last question in this group, the parent is asked to indicate in a pie-chart what percentage of the day the child hears each language.
The sixth and final question of the questionnaire is about the language abilities of the parent filling in the questionnaire. The parents are asked to self-evaluate their ability in listening, speaking, reading and writing in their mother tongue and in Dutch. This question addresses the quality of the input the child receives in each language. Since the Moroccan languages do not have an official script, the option ‘not applicable’ was added for the evaluation of reading and writing abilities. The whole questionnaire has been developed using the guidelines for questionnaires in second language research by Dörnyei (2003).
Moreover, a small pilot study with five parents in the target group was conducted. Following this, a few questions that were not understood well in the pilot study were adjusted.
2.1.3 Procedure
One of the parents or caretakers was asked to fill in the Bilingual Lexicon list about the vocabulary development of their child in both languages. The parent or caretaker was asked to tick the box in front of the words that the child produces spontaneously, but not when they are only repeated by the child. It was explained to the parent or caretaker that the words did
not have to be produced perfectly, as long as the same phonological form was always used for the same object or concept. In addition to filling in the Bilingual Lexicon list, the parent or caretaker was asked to complete the language background questionnaire.
2.2 Data collection and participants
As mentioned previously, the data are collected through parental report. The parents or caretakers were approached through the community centers of a foundation for participation, integration and emancipation called Piëzo, and through daycare centers, both in Zoetermeer.
The parents or caretakers were selected based on their language background and on the age of their child. The parents or caretakers were asked to fill in the questionnaire and the lexicon list at the community centre or the daycare centre. If this was not possible the questionnaire and the lexicon list were filled in at home and were returned to the researcher. All participants received an oral explanation of the research, the lexicon list and questionnaire. The participants could always ask the researcher questions, either at the community centre or daycare centre, or when filled in at home, by telephone.
The participants in this study are six young bilingual children divided over the four languages: two Moroccan-Arabic – Dutch bilingual children, two Tarifit-Berber – Dutch bilingual children, one Turkish – Dutch bilingual child and one Papiamentu – Dutch bilingual child as presented in table: 2.2.
Table 2.2. Participants divided over the four language groups Language
group
Moroccan- Arabic – Dutch
Tarifit-Berber – Dutch
Turkish – Dutch Papiamentu – Dutch Age in
months
20/ 42 22/ 47 35 33
Sex male/ female female/ female male female
2.3 Data analyses
The data analyses encompass a comparison of the scores in one language to the scores in the other language and both languages combined. The two single language scores, both the total scores and the total conceptual score, are plotted against the age of the children and against the percentage of the child’s language environment that each language represents.
Furthermore, a multiple regression analysis is conducted to provide insight into the contribution of the other internal and external factors, besides the quantity of input.
Moreover, the proportions of nouns, onomatopoeia and routine words, verbs, closed class words and adjectives are calculated and related to the L1 typology.
Chapter 3 Results
The outcomes of the Bilingual Lexicon lists and the language background questionnaires have been analysed. First, the scores on the Bilingual Lexicon lists are presented, then the outcomes of the language background questionnaires are discussed and the outcomes of the questionnaire are related to the scores on the Bilingual Lexicon lists.
3.1 Bilingual Lexicon lists
The scores on the Bilingual Lexicon lists for each child are presented in table 3.1. Each child has their own ID code which represents their home language, and when there is more than one child in the same language group, the number represents their age, with 1 being the younger child and 2 being the older child in the language group. This is the case for the Moroccan-Arabic language group and the Tarifit-Berber language group.
The Papiamentu and Turkish language group exist of only one child. The age of the children is presented in months. Two children are male and four children are female.
Table 3.1. The scores on the Bilingual Lexiconlists of the six young bilingual children
Child MA1 TB1 PA TU MA2 TB2
Bilingual MA-NL TB-NL PA-NL TU-NL MA-NL TB-NL
Age 20 22 33 35 42 47
Sex male female female male female female
HL score 10 34 109 85 148 186
NL score 12 111 266 89 187 193
Total score 22 145 375 174 335 379
Concept score
14 121 262 112 199 194
Lexiquotient 88 117 124 100 121 116
Result normal normal high normal normal normal
The scores in the table represent the number of words crossed off by the parent or caretaker of the child. The home language (HL) score is the score for the language other than Dutch, the Dutch (NL) score is, logically, the score for Dutch. The total score represents the scores for each language added, whereas the concept score represents the number of concepts the children reportedly have in their bilingual productive lexicon. The concept score is calculated by only counting the number of concepts the child uses, regardless of whether the child produces the word for this concept in one or in both languages. Relating this figure to the total score provides more insight into the degree of bilingualism. For MA1, 36% of the total amount of words have a translation equivalent in the other language, for TB1 this percentage is 17%, for PA it is 30%, for TU 36%, for MA2 41% and for TB2 49%. On average, for 35% of all concepts in the productive vocabulary of these bilingual children, there was a translation equivalent.
The total score is converted into a score that has been controlled for the age and language background of the children. This score is called the Lexiquotient (LQ) and makes it possible to compare children of different ages and from different language backgrounds. An LQ of 100 is the starting point to which the higher and lower scores are referenced in a norm table. This norm table also shows which scores are normal, (very) high and (very) low. When measured in both of their languages, none of these young bilingual children have very low, or
even low scores, which means there is no indication of a DLD in any of these six children.
However, if only the score for Dutch would have been taken into account and compared to the monolingual Dutch norms (Schlichting & lutje Spelberg, 2002), two of these six children (TU and TB2) would have had LQs below 80 (73 and 79 respectively) which would classify them as having (very) low scores and this would have been an indication of a DLD. When looking at the concept scores of these children this is still the case, but not when the total productive vocabulary in both languages is measured.
The Papiamentu – Dutch lexicon list has not been standardized yet and therefore the scores of PA have been compared to the norms for the Turkish – Dutch lexicon list, because this list has been standardized for the highest age. Compared to these norms, the PA participant scored above normal, but since there are no norms available for the Papiamentu – Dutch language group, it is hard to say how her score would compare to peers from the same language group. However, there is some evidence to suggest that the mother overestimated the actual productive vocabulary of her child. Therefore, PA was excluded when plotting the scores of each child against the age of the child, to see whether there is a development in productive vocabulary with increasing age. This development is presented in figure 3.1. The figure shows a growing trend in the productive vocabulary with increasing age, which is expected.
A correlation analysis was run to give more insight into the relation between the factor age and the scores in Dutch (NL), the home language (HL), the concept score and the total score. The results are presented in table 3.2. The strongest significant correlation was found between the factor age and the scores in the HL (.988), but the correlation is also significant for the factor age and the total score (.944). There are no significant correlations between the factor age and the scores in NL and the concept scores. Therefore, other factors than age must have an influence on these two scores (see paragraph 3.2). Furthermore, the
correlations between the total score, the HL score, the NL score and the concept score are all significant, which is expected.
Figure 3.1. Age plotted against the different productive vocabulary scores.
Table 3.2. The correlations between age and the productive vocabulary scores:
significant correlations (p.<.05) are marked with an asterisk.
Age Total score
HL score NL score Concept score
Age 1 .944* .988* .856 .858
Total score .944* 1 .976* .976* .967*
HL score .988* .976* 1 .906* .895*
NL score .856 .976* .906* 1 .993*
Concept score .858 .967* .895* .993* 1
As previously discussed, the semantic word categories of the Bilingual Lexicon lists can be linked to the grammatical word categories: nouns, onomatopoeia and routine words,
verbs, closed class words, adjectives and short sentences. Table 3.3 shows the proportion these grammatical classes represent as part of the total score for each child.
Table 3.3. The proportion of different grammatical word categories per child.
Child MA1 TB1 PA TU MA2 TB2
Bilingual MA-NL TB-NL PA-NL TU-NL MA-NL TB-NL
Age 20 22 33 35 42 47
Sex male female female male female female
Total score 22 145 375 174 335 379
Nouns 41% 45 % 42 % 34 % 42 % 51 %
Onomatopoeia and routine
41% 20 % 7 % 14 % 6 % 6 %
Verbs 5 % 9 % 13 % 11 % 16 % 10 %
Closed class words
9 % 10 % 19 % 22 % 19 % 16 %
Adjectives 5 % 6 % 9 % 4 % 10 % 8 %
Sentences 0 % 10 % 10 % 14 % 7 % 9 %
Nouns form the largest category for all children, however, in MA1 this category is equal to the proportion of onomatopoeia and routine words. It must be noted that a large part of the words on the Bilingual lexicon lists are nouns. This makes it less interesting to look at the proportion of grammatical word categories within one child and more interesting to compare these percentages between children of different ages and language backgrounds. Therefore, the distribution of the grammatical word categories was plotted against the age of the children in figure 3.2, 3.3 and 3.4 and against the language background of the children in figure 3.5.
Figure 3.2 shows the distribution of the grammatical word categories with increasing age. This figure shows a dichotomy in the group of children: up until 35 months of age, there is a descending trend for the categories of nouns and onomatopoeia and routine and an ascending trend for the categories of verbs, closed class words, adjectives and sentences.
However, after 35 months these trends seem to reverse. Therefore, the grammatical word categories are plotted against the age of the children in two separate figures; figure 3.3 for the ages 20-33 months and figure 3.4 for the ages 35-47 months. These trends are not only
