Measuring language development in bilingual children: Swedish-Arabic children with and without language impairment* GISELA HA˚KANSSON, EVA-KRISTINA SALAMEH, and ULRIKA NETTELBLADT
Abstract Data from ten Swedish-Arabic preschool children with language impairment and ten Swedish-Arabic children with normal language development matched for age and exposure to Swedish were analyzed. Specific tasks for both Swedish and Arabic were designed. By using the hierarchy predicted by processability theory as a yardstick, grammatical development was measured in the two languages of the children. The basic assumption is that there is a developmental sequence that all language learners follow on their way toward a target language. The results show that the bilingual children with language impairment tend to have a balanced low level of language development in both languages, whereas the bilingual children with normal language development show a higher level of language development in at least one language.
1. Introduction The number of bilingual children has increased dramatically in Sweden during the last decades. This is reflected in an increased number of referrals of bilingual preschool children to speech and language clinics for suspected language impairment (Salameh et al. 2002). Most research on child language impairment has focused on monolingual children (for a comprehensive survey, see Leonard 1998). Comparatively few studies deal with language impairment in bilingual children, although almost half of the world’s population is bilingual (de Houwer 1995), and formerly predominantly monolingual countries in Europe are becoming more multiethnic and bilingual. Since knowledge about language impairment in bilingual children is very limited, speech and language clinicians are faced with a difficult task in deciding what constitutes a language impairment and what could be attributed to a slow acquisition of a second language. Linguistics 41–2 (2003), 255–288
0024–3949/03/0041–0255 © Walter de Gruyter
256 G. Ha˚kansson et al. The diagnostic term language impairment implies that the development of language significantly lags behind development in all other areas, such as nonverbal intelligence and motor and socioemotional abilities (Nettelbladt 1983; Sahle´n and Nettelbladt 1995; Hansson 1998; Leonard 1998). Of particular interest for the present study is that children who are not monolingual speakers are usually excluded from research on language impairment and tend to be viewed as confounding data. Because of this, there is very limited knowledge of the relationship between language impairment and bilingualism. When measuring language development in a bilingual child with suspected language impairment, it should be self-evident that both languages are tested (see e.g. Damico and Hamayan 1990; Westernoff 1991; Long 1994; de Montfort Supple 1996). However, it is problematic to find suitable language tests for bilingual children in immigrant contexts with a large number of typologically different mother tongues. There are approximately 140 different mother tongues in Swedish schools. In this paper, we will investigate a new method to measure language development. An untraditional perspective will be taken. Instead of comparing the language of children with language impairment to a target norm, which is the common methodology in research on language impairment, we will establish fixed points of development for each child. In order to accomplish this we will use a theory of second-language development. In the study, data from ten Swedish-Arabic preschool children with language impairment and ten controls will be analyzed. Our aim is to explore the way in which the children with language impairment differ from the control children and also whether the levels of language development in their two languages differ. For comparing the levels, a thorough analysis of language development in both languages is required. The paper is organized in the following way. First, we will give a short overview of the two research areas, bilingualism and language impairment. Then the theoretical framework, processability theory, will be presented, together with grammatical milestones for Swedish and Arabic development. Third, the proposed stage models in both languages will be tested against data from Swedish-Arabic children. Finally, we will discuss the theoretical and clinical implications of the suggested stage models and of our results. 2. Combining different perspectives — bilingualism and language impairment This paper combines two hitherto separate research areas, viz. research on bilingual children and research on children with language impairment. We will review the literature from the respective areas below.
Measuring language development 257 2.1. Previous research on bilingual language development Language development in bilingual children has been widely discussed during the last two decades. The central issue is whether bilingual children develop one or two linguistic systems ( Volterra and Taeschner 1978). This issue has not yet been settled, partly due to methodological problems, such as pinpointing what counts as evidence (Romaine 1989) or defining what a system is ( Wode 1981). Another important issue is the age of first exposure to the target languages. De Houwer (1995) suggests that there are two different types of bilingual language acquisition. Bilingual first-language acquisition refers to the acquisition of two or more languages from birth or at most a month after birth, whereas bilingual second-language acquisition takes place after one month of age but before two years of age. After this age limit, the children are acquiring a second language. Other proposals concerning the importance of age of onset are that the age of three years marks the limit between simultaneous and successive bilingualism (MacLaughlin 1984) and that the age of six years is the critical point for acquiring native-like proficiency. However, findings from empirical research have shown that there is plenty of counterevidence to these age limits. For example, Mo¨hring (2001) found that children acquiring the French gender system with first exposure to French at the age of three were more similar to bilingual first-language learners than to L2 acquirers. Schlyter (1993) found that one of the two languages in bilingual children exposed to both languages from birth developed as an L1 and the other as an L2. Both findings contradict de Houwer’s (1995) predictions. Factors such as amount of input and individual variation are key concepts in research on bilingualism and second-language acquisition. It is a rule rather than an exception that second-language learners exhibit a great deal of variation, both within the same individual and among individuals. The variation has been dealt with, for example, in terms of motivation, cognitive style, affective factors (e.g. Schumann 1997), and linguistic context (e.g. Hyltenstam 1977).
2.2. Previous research on language impairment Earlier studies of children with language impairment have shown that the impairment is selective, that is, only specific morphological and syntactical areas are affected (for an extensive overview see Leonard 1998). Most work has been done on English children, but there is an
258 G. Ha˚kansson et al. increasing number of studies from other languages, such as German (Clahsen 1989, 1992), Swedish (Hansson and Nettelbladt 1995), Italian (Leonard et al. 1992), and Hebrew (Dromi et al. 1993). There are a number of different suggestions to explain the underlying factors behind the language impairment and since the early 1990s several theories of language impairment have been presented. They can be divided into two main groups: theories that suggest language-processing limitations as an explanation, and theories that propose an underlying deficit of the grammar of children with language impairment (for an overview see Nettelbladt 1998). Since this study is focusing on processability of language, the latter group of theories of an underlying deficit will be only briefly explained. A group of accounts claim that the core problem is a general limitation in processing capacity. The surface hypothesis, proposed by Leonard (1989), predicts that children with language impairment have difficulties in perceiving and producing elements of low phonetic substance, such as inflectional endings. The impairment is due to perceptual limitations (e.g. Leonard et al. 1992). The sparse morphology account points to the importance of the structure of the target language the child is acquiring (Leonard 1998). To test this account, data from different languages are compared. The account predicts that children with language impairment acquiring a language with sparse morphology, like English, will show more grammatical problems than children with language impairment acquiring a language with rich morphology like Italian (Leonard et al. 1992), and Hebrew (Dromi et al. 1993). An example of accounts that propose an underlying deficit is the rule deficit account (Gopnik 1990). This account claims that children with LI have difficulties in constructing implicit morphological rules. The deficits of agreement account (Clahsen 1992; Clahsen et al. 1996) claims that children with language impairment have general problems in establishing grammatical relationships. In the tense deficit account Rice and Wexler (1996) suggest that the problems are based on limitations in tensemarking.
3. Theoretical basis The idea of a fixed developmental sequence for grammatical structures has been discussed in first- and second-language acquisition research for the past three decades (e.g. Crystal et al. 1976; Meisel et al. 1981; Clahsen 1992; Pienemann 1998). One obvious implication of such a developmental
Measuring language development 259 order would be that it could give reference points to describe level of development for language learners. This approach has yielded profile measurements such as LARSP (language assessment, remediation and screening procedure; Crystal et al. 1976), grammatical profiling (Profilbo¨gen; Clahsen 1988), and rapid profile (Pienemann et al. 1988). However, with these profiles only the development of a specific target language (English and German) can be assessed. There is a need for a more general framework for the analysis of other languages as well. 3.1. Processability theory As the theoretical framework for the study, Pienemann’s processability theory (PT ) will be used (Pienemann 1998; Pienemann and Ha˚kansson 1999). This is based on Levelt’s (1989) model of language production and it is non–language-specific. Processability theory focuses on the dynamic character of language acquisition and spells out in detail the processing prerequisites that are needed for the automatization of grammatical rules on different developmental levels. In other words, the task of acquiring a language includes the acquisition of the procedural skills needed for the processing of the language. It follows from this that the sequence in which the target language (TL) unfolds in the learner is determined by the developmental sequence of processing routines that are needed to handle the TL’s components (Pienemann and Ha˚kansson 1999: 385).
By choosing a psycholinguistic framework of language development for children with language impairment, we want to stress that we regard grammar from a learning perspective. The children are treated as language learners, building their own interlanguage grammar, and not as having a static, defective variety of the target language. In other words, the language production in these children is not measured against target norms. We are rather trying to establish fixed points of development in order to compare the two languages. The precise procedures needed for the processing of each stage in the development are described in Table 1. As a first step in this developmental route the learner identifies and acquires the words of the target language ( level 1). The next step is to categorize the lexicon and list the diacritic features of the lexemes in the lexicon. This is the level of lexical morphology ( level 2). Lexical morphology is a necessary prerequisite for phrasal morphology ( level 3) to be processable. The processing of phrasal morphology allows the learner to unify the features of head and modifier in a phrase. When phrasal morphology is automatized, interphrasal morphology is processable
260 G. Ha˚kansson et al. Table 1. Hierarchy of processing procedures (from Pienemann and Ha˚kansson 1999) Processing procedures
Structural outcome
5. 4. 3. 2. 1.
main and subclause interphrasal information exchange/merging phrasal information exchange/merging lexical morphemes words
Subordinate-clause procedure S-procedure Phrasal procedure Category procedure Word or lemma access
( level 4). This step implies that the grammatical functions of the words in a clause will be accessible and feature unification between phrases is possible. At this level the rule that regulates subject–verb inversion is processable. Finally, when main-clause word-order rules are automatized, the hierarchical relation between main and subordinate clauses is processable and the learner can apply different grammatical rules in main and subordinate clauses. The processes mentioned above are at a very general level and they are realized in different ways in typologically different languages. We will now proceed to the developmental hierachies predicted for Swedish and Arabic.
4. Grammatical structures in Swedish and Arabic 4.1. Introduction The sequences in which the target language develops are described in a non–language-specific manner in the section above. Now we will take a closer look at Swedish and Arabic grammar in terms of processing complexity. Swedish, a Germanic language, is characterized as an analytic language with agglutinating and synthetic features. Arabic belongs to the Semitic language family and is a synthetic language with fusional features. Both languages encode functions such as gender, number, and definiteness by morphological markings and use morphology at the lexical as well as the phrasal and interphrasal levels. We will only describe those parts of morphology and syntax that are relevant for processability theory, and we will discuss the order in which they are predicted to develop during acquisition. The key structures for Swedish were developed in great detail in Pienemann and Ha˚kansson (1999) and we will be adopting their framework here. For Arabic, we are relying mainly on existing descriptions of Modern Standard Arabic (Holes 1995) and also on earlier
Measuring language development 261 accounts of the processability of Arabic grammar (Mansouri 1995, 1999, 2000). A major problem in this context, since we are dealing with preschool children learning a language in exile, is the lack of descriptions of colloquial Arabic (Eid 1990). There are large differences between Modern Standard Arabic and colloquial varieties of Arabic. In fact, Arabic is a classic example of language in a diglossic situation (Ferguson 1959). This term refers to a situation where different varieties of a language co-occur throughout a speech community, each with a distinct social function. Both varieties are standardized to some degree, but one of the varieties is considered to have higher social prestige and have a more formal character. The high variety is typically learned in school and used in formal contexts. The colloquial varieties are characterized by a simplified morphology and syntax. Since PT deals with morphology and syntax, this has consequences for the proposed stages of development in this study. There are at least two complications for our study. We have to base our stage model on Modern Standard Arabic, a language that preschool children naturally have not yet been exposed to in school. Second, we are dealing with a language in exile, which means that different varieties of Arabic are mixed. It is important to keep in mind, however, that the children with language impairment are not compared to a target-language norm. The base-line data are from the control group, children exposed to the same language in exile. In setting up the test structures we had to find minimal pairs in obligatory contexts. This means that our choice of grammatical structures in the description of individual grammars is somewhat constrained.
4.2. Swedish and Arabic morphology 4.2.1. Level 1 — invariant forms. At this level the learner only produces lexical items that are not inflected; that is, they are used in invariant forms. 4.2.2. Level 2 — lexical morphology. Lexical morphology involves the addition of a morpheme to a stem. In Swedish, suffixes are added to nouns to indicate definiteness and number and to verbs to indicate tense. In Arabic, plural can also be expressed by adding a suffix to the noun. Arabic has different ways to express plural. The regular plural is very much like Swedish (and English), simply adding a suffix to the noun without necessitating internal changes in the word. The irregular plural (broken plural ) requires changes within the word itself (infixes) and there are at least six different classes (e.g. walad–awlaad ‘boy–boys’).
262 G. Ha˚kansson et al. The examples below illustrate regular plural morphology in Swedish and in Arabic. The regular plural is expected to appear in both languages at level 2. (1) (2)
katt cat.SING.UTER bisse cat.SING.FEM
katt - er (Swedish) cat.PLUR.UTER biss - aat (Arabic) cat.PLUR.FEM
Verbal morphology is different in the two languages. Swedish verbs can be marked with suffixes for tense (and finiteness, see discussion in Platzack 1992; Ha˚kansson 2001). Swedish tense marking is quite consistent and salient. There are three different weak conjugations, one class of strong verbs and one class of irregular verbs. The suffixes in the different conjugations have slightly different, phonologically determined, forms for present (stem+-r or stem+-er) and past tense (stem+-de, or stem+-te). Strong verbs typically exhibit vowel shift (‘‘umlaut’’) and irregular verbs have idiosyncratic forms. In contrast to Swedish, Arabic usually expresses tense by vocalic changes, or infixes. The consonants carry the primary meaning and the vocalic patterns mark categories such as tense and aspect. ( We will not account for the differences between deictic and nondeictic time reference in this paper, nor for any tense–aspect differences. For the sake of simplicity we will use the terms ‘‘present’’ to indicate ‘‘nonpast’’ and ‘‘past’’ to indicate ‘‘perfect’’ in Arabic.) (3)
(4)
Tense ‘‘nonpast’’ lek - er play.PRES ( j)ishrab 3P.MASC.drink.SING.PRES
‘‘past’’ lek - te (Swedish) play.PAST shirib (Arabic) 3P.MASC.drink.SING.PAST
When it comes to other verbal suffixes there is another striking difference between Swedish and Arabic (and also between Arabic and English). Arabic has obligatory person markings on the verb. However, like for instance Italian and Spanish, it is a pro-drop language, where pronominal subjects are optional. This implies that the verb form in itself may include a subject marker (as in example [4] above). The Arabic morphology expressing singular and plural person is illustrated in examples (5) and (6), below. Singular is the unmarked case (with a zero morpheme %) and plural is expressed by the suffix -uu.
Measuring language development 263 (5) (6)
( j) - i 3P.MASC ( j) - i 3P.MASC
shrab % drink.SING.PRES shrab- uu drink.PLUR.PRES
haliib milk haliib milk
It is difficult to decide whether the morpheme expressing person belongs to the lexical, phrasal, or interphrasal level. When there is agreement between subject and verb it is natural to consider this as interphrasal agreement. For the present study, we will assume the Arabic verb suffix to have two different interpretations. First, it is interpreted as part of the verb, a lexical suffix, at the early level when the learners do not process clauses, but only words. At this early level, the verb marking in itself does not involve any exchange of grammatical information between constituents. However, the full agreement pattern with agreement between the subject and the verb will involve considerable grammatical merging of information and we will therefore place this structure at the interphrasal level. 4.2.3. Level 3 — phrasal morphology. In Swedish, expanded noun phrases involve full agreement between modifiers and head noun. The article and adjective agree with the head noun in gender (neuter or uter), definiteness (definite or indefinite) and number (singular or plural ). In other words, there is exchange/merging of grammatical information between article, adjective, and noun. There are some restrictions in the system; for example, definite adjectives do not agree in gender and number with the noun, and the plural definite article does not agree in gender with the noun. In Arabic there is full agreement in noun phrases. The modifiers agree with the head noun in gender (masculine or feminine) and definiteness (plus or minus definite). If the head noun is human, they also agree in number (singular, dual, or plural ). Adjectives behave like nouns and take the same prefix for definiteness and gender as the noun. Examples (7)–(10) below illustrate the Swedish and Arabic structures. (7)
(8)
(9)
den stor-a pojke-n DEF.UTER big.DEF boy.DEF.UTER.SING. ‘the big boy’ de stor-a pojk-ar-na DEF.PLUR. big.DEF boy.DEF.UTER.PLUR ‘the big boys’ albint alhilw-e DEF.ART girl.FEM.SING DEF.ART pretty.FEM.SING ‘the pretty girl’
264 G. Ha˚kansson et al. (10)
albanaat alhilw-aat DEF.ART girl.FEM.PLUR DEF.ART pretty.FEM.PLUR ‘the pretty girls’
4.2.4. Level 4 — interphrasal morphology. At the level of interphrasal morphology, the different grammatical functions of the constituents in the clause are identified. Here, we find exchange, or merging, of information between NP and VP. In Swedish the processing of this level is realized in subject–verb inversion and predicative agreement. This inversion is obligatory when there are fronted elements, that is in wh questions where the subject is not questioned, and in topicalized declaratives. The verb is always in the second position in main-clause declaratives. The rules of Swedish word order are illustrated below in example (11), a declarative clause, and (12), a topicalized declarative clause with subject–verb inversion. Swedish nouns agree in number and gender with the adjective in predicative position (example [13]). (11) Han ko¨pte en bok iga˚r he.buy.PAST.ART.book. yesterday ‘he bought a book yesterday’ (12) Iga˚r ko¨pte han en bok yesterday.buy.PAST.he.ART.book ‘yesterday he bought a book’ (13) Pojkarna a¨r stora boy.PLUR.DEF COP big.PLUR ‘the boys are big’ In Arabic, interphrasal agreement is realized in subject–verb agreement. As was mentioned above, different proposals have been made to account for its complexity (see e.g. Eid 1990; Mohammad 1990; Bolotin 1995). The agreement between the subject and the verb depends on the surface order of the subject and the verb. In other words, it involves syntax as well as morphology. Due to the pro-drop phenomenon, which means that pronominal subjects are not obligatory, many clauses in Arabic lack subjects. The optional character of overt subjects makes it difficult to test the structure. Examples of Arabic subject–verb agreement are given in examples (14) and (15). (14)
alwalad ( j)i-shrab haliib DEF.ART boy.SING 3P.MASC.drink.SING milk ‘the boy is drinking milk’
Measuring language development 265 (15)
alawlaad ( j)i-shrab-uu haliib DEF.ART boy.PLUR 3P.MASK.drink.PLUR milk ‘the boys are drinking milk’
4.2.5. Level 5 — subordinate-clause procedure. In Swedish, the difference between main clauses and subordinate clauses is expressed by a word-order opposition. The verb occupies second position in main clauses but has a later position in subordinate clauses. In subordinate clauses the word order is always SVO (COMP–SVO). The negator is postverbal in main clauses, but preverbal in subordinate clauses. The relative pronoun is som, an invariant form. (15)
Flickan gungar inte girl.DEF swing.PRES not ‘the girl swings not’ (16) Ha¨r a¨r flickan som inte gungar here COP girl.DEF who not swing.PRES ‘this is the girl who does not swing’ In Arabic there is no word-order opposition that differentiates between main and subordinate clauses. One possible structure that may differentiate main clauses from subordinate clauses is the use of subjunctive instead of indicative on verbs in subordinate clauses. Since subjunctive is not obligatory in colloquial Arabic we have chosen to disregard this structure in the present study.1 Summarizing, the processability hierarchies for both Swedish and Arabic are given in Table 2.
Table 2. Hierarchies with selected structures for Swedish and Arabic development Processing prerequisites
Outcome Swedish grammar
5. 4. 3. 2.
subordinate clause word order V2, pred agreement NP, VP agreement verb suffix: present, past noun suffix: plural
clause boundary S-procedure phrasal category category procedure
1. word/lemma
‘‘words’’
Outcome Arabic grammar
subject–verb agreement NP agreement verb suffix: person verb inflection: present, past noun suffix: plural ‘‘words’’
266 G. Ha˚kansson et al. 5. Earlier studies 5.1. Acquisition of Swedish morphosyntax Studies on second-language (L2) learners acquiring Swedish grammar show that this group of learners follows the predicted route closely. There is a large body of studies of the acquisition of Swedish grammar in L2 learners involving more than a thousand informants (see Pienemann and Ha˚kansson 1999 for an overview). Both adult L2 learners (e.g. Hyltenstam 1977; Bolander 1988) and child L2 learners (Andersson 1992; Ha˚kansson and Nettelbladt 1993, 1996; Salameh et al. 1996) are investigated and there are longitudinal as well as cross-sectional studies. The subject–verb inversion ( level 4) is the structure that has attracted most interest. Usually learners of Swedish start out with a SV word order ( level 2), then advance to preposing the adverb ( level 3) without changing the order subject–verb before they proceed to invert subject–verb in topicalized clauses (Ha˚kansson and Nettelbladt 1993, 1996; Hyltenstam 1977; Bolander 1988). This confirms the predictions for syntactic development at levels 2, 3, and 4. The studies on the acquisition of Swedish syntax give a much more coherent picture of a developmental route than the studies on agreement morphology. There are, however, some studies on L2 Swedish comparing emergence of morphology and emergence of syntax. In Ha˚kansson (2001) lexical morphology on verbs ( level 2) emerged before inversion ( level 4). In a cross-sectional study Salameh et al. (1996) investigated noun-phrase agreement ( level 3) and inversion in questions ( level 4) in twenty Arabicspeaking children acquiring Swedish as a second language. An implicational order was found between these structures. Another study on morphology is Glahn et al. (2001), where the acquisition of Danish, Norwegian, and Swedish in university students was investigated. Nounphrase agreement ( level 2) was found to occur before predicative agreement ( level 4), which came before word order in subordinate clauses ( level 5). However, Glahn et al. also found other factors that influenced the developing learner grammar; for example, there was a difference between number agreement and gender agreement. This highlights the problem that choice of structures may be crucial to the result. Interestingly enough, the Swedish structures develop in exactly the same order in bilingual children as in monolingual children with language impairment (Ha˚kansson and Nettelbladt 1993, 1996). However, monolingual children without language impairment have no period of development between levels 2 and 4. In the production from these children lexical morphology and inversion emerge at the same time, before the age of
Measuring language development 267 two years (Santelmann 1995; Platzack 1996). The reason behind this has not been fully explained, but is is possible that the one-word stage and two-word stage help the children organize their grammar. They may develop grammar and lexicon simultaneously. As mentioned above, the age factor is often discussed in research on bilingual language development. It has been suggested that the age of first exposure to the target language (age of onset, AO) is a predictor of the development. De Houwer suggests that bilingual first-language acquisition has to start before the age of one month, bilingual secondlanguage acquisition before the age of two years, and then secondlanguage acquisition takes over. MacLaughlin (1984) proposes the age of three years, whereas Long (1993) claims that children with an age of onset between zero and six years are the ones likely to reach native-like proficiency. The bilingual Swedish-Arabic children in the present study are exposed to Arabic from birth and to Swedish at a later age (AO 1:6–5:6). Arabic is used at home and Swedish in preschool. It is possible that Swedish is acquired as an L1 among the younger children and as an L2 in the older children. However, this is hard to decide without a more detailed study of the interaction and input at the day-care centers.
5.2. Acquisition of Arabic morphosyntax There are two studies on L2 acquisition of Modern Standard Arabic from a processability perspective, both involving students studying Arabic at university level. Lykke-Nielsen (1996) focused on the acquisition of Arabic noun phrases in one Danish learner. The predictions were only partly confirmed, and Lykke-Nielsen suggested that there is a distinction between progressive or regressive transfer of information between the constituents in the noun phrases. This complicates the picture of phrasal agreement. In Mansouri (1995, 1999, 2000) the acquisition of Arabic was investigated in a longitudinal study of four learners. The aim was to test five proposed stages in the processability hierarchy. The results show that, although there was a lot of interlearner variability, the hierarchy was followed and the predictions were largely confirmed. As in the case of L2 acquisition of Swedish, however, the picture was clearer for syntax than for morphology. Mansouri found that lexical morphology (for example definite markers, semantic gender, and past tense) emerged before phrasal morphology (NP agreement). Mansouri also analyzed the different steps in subject–verb agreement marking. The agreement marking developed from reduced agreement to full agreement. Finally, the learners were able to distinguish between different clause types and use
268 G. Ha˚kansson et al. different types of agreement in different clause types. One major difference between Mansouri’s study and our own is that his informants produced a large number of clauses with full subjects, which our subjects very rarely used (only two examples in all of the material ). This is probably due to the fact that in Mansouri’s study the informants were university students acquiring Arabic in a formal setting and using a more context-free language. There are very few studies of L1 acquisition of Arabic, the most important exception being the study by Omar (1973). She found that noun inflection for plural and gender was acquired at the age of three years and the phrasal agreement was acquired about half a year later. These findings are in accordance with our predictions namely that lexical morphology is processable before morphology marking phrasal agreement. However, the other structures investigated in her study, dual inflection and irregular inflections, fall outside the PT hierarchy. There are no examples of the acquisition of later levels.
6. Aims of the study The aims are presented as two hypotheses. Hypothesis 1. The children will develop grammatical structures for their two languages in the implicational order suggested above. Hypothesis 2. The children in the control group have developed at least one of their two languages to a high level (4 or 5 for Swedish, and 3 for Arabic). The children with language impairment have a lower level of development ( level 2), and they will have a lower level in both languages.
7. Method 7.1. Participants and data Ten bilingual Swedish-Arabic children with language impairment were chosen as subjects for the study. Permission was granted by the Ethics Committee of the Medical Faculty of the University of Lund. Inclusionary criteria were that the children had recieved a diagnosis of severe language impairment by a speech and language clinician at the Speech and Language Clinic at University Hospital in Malmo¨, were within the age span of 4–7 years, and lived in an ethnically diverse area.
Measuring language development 269 All children displayed grammatical problems and were well below age expectations in both languages. Only one child had additional severe phonological problems. The children were included after written consent from their parents. In order to attempt to validate the diagnosis of severe language impairment that had been assigned to these children, a design with a matched control group was chosen. MLU (mean length of utterance) matches are widely used within research on language impairment. However, MLU matching is not used within second-language research. This raises interesting issues about MLU controls in two languages, both Swedish and Arabic. Some of the children in the study have been exposed to Swedish for less than a year and are very limited in their production of Swedish, which implies that the MLU controls have to be very young. It is doubtful whether a comparison between a five-year-old bilingual child and a 18-month-old monolingual child is meaningful since there are large differences in their cognitive as well as their communicative abilities. Instead of MLU match, age matching was used. Ten bilingual Swedish-Arabic children, with normal language development in both languages according to parental report and bilingual and Swedish preschool staff, were chosen as a control group. The two groups of children were matched according to age, gender, time of exposure to Swedish in preschool, and original Arabic dialect of the parents. The groups were if possible matched according to educational level of parents, where 1 denoted university level, 2 upper secondary level, and 3 compulsory school. All children were born in Sweden, and the parents of the children with language impairment had lived in Sweden 6–16 years with a mean of 11 years. Corresponding figures for the parents of the children in the control group were 5–15 years with a mean of 9 years. All children lived in ethnically diverse areas in Malmo¨. The majority of parents spoke Lebanese, Palestinian, and Syrian dialects of Arabic. According to Fischer and Jastrow (1980) these three dialects are considered to belong to the same group. The Iraqi dialect, spoken by parents of three children, and the Gulf dialect, spoken by parents of one child, both belong to different groups. Consideration must be given, however, to the fact that these Swedish-Arabic children are exposed daily to both the Syrian-Lebanese-Palestinian and the Iraqi dialects from other children and bilingual staff. 7.2. Pretesting Each child was pretested with a production test and a comprehension test in Swedish. The Lund Test of Phonology and Grammar (Holmberg
270 G. Ha˚kansson et al. and Stenkvist 1983) is a production test that includes plural forms and genitive markings of nouns, comparison markers on adjectives, tense markers on verbs, pronouns, prepositions, and the placement of negation. The children with language impairment had low scores with a mean of 16.1%, although one child, LI:7 Abir, managed to score 57.1% of the points. The mean for the children in the control group was 49%. The Swedish Test of Language Comprehension, SIT (Hellquist 1989) was used for comprehension. This test includes different forms of verb tenses, pronouns, prepositions, conjunctions, and negators. All children performed better on the comprehension test. The mean for the children with language impairment was 55.6%, and for the children in the control group 72%. The Mann–Whitney U-test showed a significant differerence between the two groups on the comprehension test (p=0.043) but not the production test (p=0.075). The factors that affected the result are age and time of exposure to Swedish. These factors did not differ between the groups. When a parametric analysis of variance was made and age and exposure were adjusted for, there was a significant difference between the groups on both the comprehension test (p=0.026) and the production test (p=0.006). Both groups were tested with Leiter International Performance Scale, standardized in Swedish (Leissner et al. 1962). All children also passed both a hearing screening (20 dB at 500, 1000, 2000, and 4000 Hz) and an oral-motor screening (Holmberg and Bergstro¨m 1996). Table 3 displays an overview of the children including test scores from the pretesting. In Table 4 the children are compared as groups. Six boys and four girls had language impairment, while there were four boys and six girls in the control group. The children in the control group were also slightly younger as a group and had spent less time in a Swedish preschool. Parental levels of education differed only slightly. It is important to point out that the standardization of the Leiter scale is made for Swedish monolingual children, yet it is used in this study to achieve a measure, roughly comparable in terms of general cognitive development. The children were compared as groups, not as individuals. The results show that the difference between the two groups is 1 point. The mean for Swedish monolingual children is 5 stanine points+2SD.
7.3. Procedure 7.3.1. Pilot tests. Structures that were regarded as relevant for the different levels of processability were selected, and tests in Swedish and
Key:
Lebanese Iraqi Gulf-dialect Lebanese Lebanese Iraqi Lebanese Lebanese Lebanese Syrian
Iraqi Lebanese Palestinian Lebanese Lebanese Lebanese Lebanese Lebanese Lebanese Lebanese
Arabic dialect
3:7 2:10 1:6 2:8 0:6 2:11 1:11 1:10 0:6 1:2
2:1 1:11 1:6 0:7 3:10 3:7 2:10 2:6 3:0 0:10
Time in preschool
LI denotes language impairment and ND normal development.
5:8 4:9 5:8 6:7 5:1 5:3 4:5 4:2 3:11 4:1
F F F M F M M F F M
Control group ND:1 Hoda ND:2 Nadja ND:3 Sannah ND:4 Yasin ND:5 Zahla ND:6 Karim ND:7 Mansour ND:8 Azra ND:9 Inaaz ND:10 Adil
Age
impairment M 4:7 M 4:4 M 3:11 M 6:2 F 6:7 M 5:8 F 6:1 F 4:8 M 4:5 F 3:10
Sex
3 1 2 3 3 2 3 3 1 1
1 3 3 3 2 2 3 2 2 3
Educational level parents
Swedish-Arabic children with language impairment and in the control group
Children with language LI:1 Zuhair LI:2 Anwar LI:3 Taha LI:4 Mahmoud LI:5 Hanin LI:6 Mohsen LI:7 Abir LI:8 Suha LI:9 Abdul LI:10 Manal
Table 3.
77.6 79.6 73.5 61.2 10.2 83.7 59.2 34.7 4.1 6.1
16.3 2 6.1 18.4 14.3 10.2 57.1 26.5 12.2 0
(38/49) (39/49) (36/49) (30/49) (5/49) (41/49) (29/49) (17/49) (2/49) (3/49)
(8/49) (1/49) (3/49) (9/49) (7/49) (5/49) (28/49) (13/49) (6/49) (0/49)
Lund test % (ratio)
89.1 89.1 93.6 89.1 58.7 91.3 82.6 60.9 23.9 41.3
58.7 54.3 34.8 58.7 89.1 54.3 71.7 60.9 52.3 21.7
SIT %
(41/46) (41/46) (43/46) (41/46) (27/46) (42/46) (38/46) (28/46) (11/46) (19/46)
(27/46) (25/46) (16/46) (27/46) (41/46) (25/46) (33/46) (28/46) (24/46) (10/46)
(ratio)
Measuring language development 271
272 G. Ha˚kansson et al. Table 4. A comparison between the groups with LI and the control group
Children with language impairment Children in control group
Sex
Mean age
Mean time in preschool
Mean education level
Mean Leiter stanine
6 4 4 6
5.03 span 3.19–6.7 4.96 span 3.11–6.7
2.27 span 0.7–3.10 1.86 span 0.6–3.7
2.4
4.9
2.2
5.9
m f m f
Arabic were designed to create obligatory contexts for these structures. A pilot test was administered to study the childen’s response to the chosen structures. A second pilot test in Swedish was administered to check elicitation strategies, and the results showed that children with language impairment were clearly hampered in their performance by extensive prompting (Salameh 2000). As a consequence a strict test protocol was developed for both languages. 7.3.2. Final test versions. In the final versions of the Swedish and Arabic tests, different structures were chosen in order to create more contexts for some of the levels. For example, for level 2, plural and past tense were chosen for Swedish and plural, person, and past for Arabic. (Since past tense is expressed by a fused inflection, this means that we tested both suffixes and fusion in Arabic.) For level 3, agreement between adjective and noun and agreement between auxiliary and main verb were tested in Swedish. Only noun-phrase agreement was tested in Arabic. For level 4 in Swedish, we chose both predicative agreement and subject–verb inversion. For Swedish level 5, placement of negation was chosen. No test structures for level 4 and 5 in Arabic were identified, since the structures on these levels are either optional or absent in colloquial Arabic. The structures are summarized in Table 5.
Table 5. Elicited structures PT levels
2
3
4
5
Swedish
number tense
NP agr VP agr
V2 pred agr
subclause wo
Arabic
person number tense
NP agr
no structure
no structure
Measuring language development 273 7.3.3. Test procedure. The second author tested all children in both Swedish and Arabic in order to obtain a coherent test procedure. In both languages the structures (see Table 5) on levels 2 and 3 were elicited by the use of toys and pictures. The same pictures and toys were used for NP agreement on level 3 and predicative agreement items on level 4 in order to avoid lexical problems. Word order ( V2-rule) in Swedish on level 4 was elicited by two short sets of picture sequences. To elicit relative clauses on level 5 a ‘‘Lotto game’’ was used (Ha˚kansson and Hansson 2000). Negated main clauses were first elicited from the child and then negated subclauses. Four or five sessions (about 45 minutes each, including pretesting) for each child were required, altogether 3–4 hours. Most children were tested in their day-care center. One boy was tested in the speech clinic and one in his home. The whole procedure was tape-recorded, using a Sony TCS 580V. 7.3.4. Scoring. Number of occurrences in obligatory contexts were calculated. An emergence criterion was used. If the child used a systematic contrast, for example the same verb in two forms, in both present and past tense, one example of each tense was enough for the child to score on lexical morphology. If the verb was only used in the same invariant form all the time, no points were given, even if the verb happened to be inflected according to the target structure in one particular context. As was mentioned earlier the developmental perspective allows us to consider all plural markings equal if they involve systematic contrasts. For example, if the child chooses the wrong suffix for a certain word, we still regard it as a marker of plural (cf. the English minimal pair child– childs instead of child–children). According to processability theory, finding out which suffixes go with individual nouns is another acquisitional task, which has nothing to do with processing but belongs to the lexicon. The assessment of test results was made by the first and second author and also by a native speaker of Arabic. 8. Results The results will be presented level by level. An overview of the results on all levels is shown in Table 10. 8.1. Level 1 — word level All children with language impairment produced words in Swedish as well as Arabic, although some children only produced invariant forms.
274 G. Ha˚kansson et al. One example of an invariant form in Arabic is when the word bisse ‘cat’ is produced in this form in both singular and plural contexts. No invariant forms were found among the children in the control group.
8.2. Level 2 — lexical morphology Table 6 shows the results on this level for the different structures in Swedish and Arabic for the children with language impairment and the control group. Among the children with language impairment, five used lexical morphology in Swedish, and six in Arabic. Three of the children did not use morphology in either language. In Arabic four children produced both person and plural markings while one child produced only person markings. One child produced only plural markings. No child in this group produced tense. All children from the control group used lexical morphology, two of them only in Arabic. Three children used all three structures in Arabic, Table 6. The test result of the different structures on level 2 in Swedish and Arabic Level 2 Swedish plur/tense
Level 2 Arabic pers/plur/tense
Children with language impairment LI:1 Zuhair −/− LI:2 Anwar −/− LI:3 Taha −/− LI:4 Mahmoud −/− LI:5 Hanin +/− LI:6 Mohsen +/− LI:7 Abir +/+ LI:8 Suha −/+ LI:9 Abdul +/− LI:10 Manal −/−
−/−/− −/+/− −/−/− +/+/− +/+/− +/+/− +/+/− −/−/− +/−/− −/−/−
Control group ND:1 Hoda ND:2 Nadja ND:3 Sannah ND:4 Yasin ND:5 Zahla ND:6 Karim ND:7 Mansour ND:8 Azra ND:9 Inaaz ND:10 Adil
+/+/+ +/+/+ +/+/+ +/−/− +/+/− +/+/− +/+/− +/+/− +/+/− +/+/−
+/+ +/+ +/− +/+ −/− +/+ +/+ +/− −/− +/−
Measuring language development 275 six used both person and plural markings, and one child only person markings. Below are examples of structures produced in Arabic on this level by children with language impairment: (17)
Tester: Child LI:7 Abir
Tester: Child LI:7 Abir
Shu ‘amjamal al-walad? ‘What is the boy doing?’ ashlab (ashrab) drink.PRES.MASC.3P.SING ‘(he) drinks’ Shu ‘amjamaluu al-awlaad? ‘What are the boys doing?’ ashlab-uu (ashrabuu) drink.PRES-MASC-3P-PLUR
haliib milk
haliib milk
Child LI:7 Abir managed to produce person markings. Child LI:3 Taha below made phonological simplifications in his pronounciation of the word ‘ashrab and only produced invariant forms. (18)
Tester:
Shu ‘amjamal al-walad? ‘What is the boy doing?’ Child LI:3 Taha atlab (ashrab) drink.PRES.MASC.3P.SING Tester: Shu ‘amjamaluu al-awlaad? ‘What are the boys doing?’ Child LI:3 Taha atlab (ashrab) drink.PRES.MASC.3P.SING
maj water
maj water
8.3. Level 3 — phrasal level As is shown in Table 7, only one of the children with language impairment, LI:7 Abir, used phrasal morphology in Swedish, but not in Arabic. No child with language impairment produced phrasal morphology in Arabic. No child with language impairment used phrasal morphology in both Swedish and Arabic. Five of the children in the control group produced phrasal morphology in Swedish and seven in Arabic. Three children used phrasal morphology in both Swedish and Arabic. 8.4. Level 4 — interphrasal morphology Only one child with language impairment used interphrasal morphology in Swedish, LI:9 Abdul; see Table 8. None of the children with language impairment produced interphrasal morphology in Arabic (subject–verb
276 G. Ha˚kansson et al. Table 7. The test result of the different structures on level 3 in Swedish and Arabic Level 3 Swedish NP-agreement VP-agreement
Level 3 Arabic NP-agreement
Children with language impairment LI:1 Zuhair −/− LI:2 Anwar −/− LI:3 Taha −/− LI:4 Mahmoud −/− LI:5 Hanin −/− LI:6 Mohsen −/− LI:7 Abir −/+ LI:8 Suha −/− LI:9 Abdul −/− LI:10 Manal −/−
− − − − − − − − − −
Control group ND:1 Hoda ND:2 Nadja ND:3 Sannah ND:4 Yasin ND:5 Zahla ND:6 Karim ND:7 Mansour ND:8 Azra ND:9 Inaaz ND:10 Adil
− + + − + + + + + −
−/+ +/+ +/− −/+ −/− +/+ −/− −/− −/− +/−
agreement is optional in colloquial Arabic; consequently there were no test structures for this level ). Six children in the control group produced structures from this level in Swedish. One of these children, ND:2 Nadja, produced the optional structure subject–verb agreement of level 4 in Arabic, while answering an Arabic test item from level 3: (19)
Child ND:2 Nadja ana akhoud al- hamra I.1PERS.SING take.PRES.1P.SING DEF red.FEM.SING ‘I take the red one’ Child ND:2 Nadja inti t-akhdi asyou.2PERS.FEM take.PRES.2.FEM.SING DEF safra yellow.FEM.SING ‘you take the yellow one’
Measuring language development 277 Table 8. The test result of the different structures on level 4 in Swedish Level 4 Swedish V2/ pred agreement Children with language impairment LI:1 Zuhair LI:2 Anwar LI:3 Taha LI:4 Mahmoud LI:5 Hanin LI:6 Mohsen LI:7 Abir LI:8 Suha LI:9 Abdul LI:10 Manal
−/− −/− −/− −/− −/− −/− −/− −/− −/+ −/−
Control group ND:1 Hoda ND:2 Nadja ND:3 Sannah ND:4 Yasin ND:5 Zahla ND:6 Karim ND:7 Mansour ND:8 Azra ND:9 Inaaz ND:10 Adil
+/− +/+ +/+ +/+ −/− +/+ +/+ −/− −/− −/−
8.5. Level 5 — subordinate-clause level None of the children with language impairment produced any structures from level 5, as can be seen in Table 9. Four children in the control group produced subordinate clauses in Swedish. As earlier mentioned, this level was not tested in Arabic.
8.6. Implicational order Table 10 shows the implicational order for both groups and all levels in Swedish and Arabic. As the table shows, the tested grammatical structures in Swedish have developed in an implicational order. The scalability is 0.86 and 0.94 respectively (the coefficient of scalability should be above 0.60 before scalability can be claimed). All children that produced phrasal morphol-
278 G. Ha˚kansson et al. Table 9. The test result of the different structures on level 5 in Swedish Level 5 Swedish Subclause wo Children with language impairment LI:1 Zuhair LI:2 Anwar LI:3 Taha LI:4 Mahmoud LI:5 Hanin LI:6 Mohsen LI:7 Abir LI:8 Suha LI:9 Abdul LI:10 Manal
− − − − − − − − − −
Control group ND:1 Hoda ND:2 Nadja ND:3 Sannah ND:4 Yasin ND:5 Zahla ND:6 Karim ND:7 Mansour ND:8 Azra ND:9 Inaaz ND:10 Adil
+ + + + − − − − − −
ogy also used lexical morphology. Two children who produced structures on level 4 did not show any examples of phrasal morphology. The four children that produced subordinate clauses in Swedish used structures on all levels. Also in Arabic, the order is implicational, although there are only three levels. No child is producing an Arabic structure to the right in the table without also producing a structure to the left.
8.7. The hypotheses The results from the tests show that the structures develop in the implicational order as predicted for both languages for eight out of ten children with language impairment and nine out of ten children without language impairment. Our first hypothesis, that the structures would develop in the suggested order, is thus confirmed. Our second hypothesis predicted that children in the control group
Measuring language development 279 Table 10. Implication table for Swedish (scalability 0.86 for L1 and 0.94 for ND) and Arabic (scalability 1.0 in both groups) Level Swedish 1 2 3
4
5
LI:1 Zuhair LI:2 Anwar LI:3 Taha LI:4 Mahmoud LI:10 Manal LI:5 Hanin LI:6 Mohsen LI:8 Suha LI:7 Abir LI:9 Abdul
+ + + + + + + + + +
− − − − − + + + + +
− − − − − − − − + −
− − − − − − − − − +
− − − − − − − − − −
ND:5 Zahla ND:9 Inaaz ND:10 Adil ND:8 Azra ND:7 Mansour ND:6 Karim ND:1 Hoda ND:2 Nadja ND:3 Sannah ND:4 Yasin
+ + + + + + + + + +
− − + + + + + + + +
− − − − − + + + + +
− − − − + + + + + +
− − − − − − + + + +
Level Arabic 1 2 3
4
5
LI:1 Zuhair LI:3 Taha LI:10 Manal LI:2 Anwar LI:4 Mahmoud LI:5 Hanin LI:7 Abir LI:8 Suha LI:9 Abdul LI:6 Mohsen
+ + + + + + + + + +
− − − + + + + + + +
− − + − − − − − − −
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
ND1: Hoda ND:4 Yasin ND:10 Adil ND:3 Sannah ND: 5 Zahla ND:6 Karim ND:7 Mansour ND:8 Azra ND:9 Inaaz ND:2 Nadja
+ + + + + + + + + +
+ + + + + + + + + +
− − − + + + + + + +
0 0 0 0 0 0 0 0 0 +
0 0 0 0 0 0 0 0 0 0
have developed at least one language at a high level, while children with language impairment have a low level in both languages. Table 11 shows a comparison of levels in Swedish and Arabic. The results confirm the second hypothesis with some exceptions. Among the children with language impairment eight out of ten children produce structures on levels 1 and 2 only, that is, they have a low level in both languages. One child, LI:7 Abir, produces structures on level 3 in Swedish and level 2 in Arabic. Another child, LI:9 Abdul produces level 4 in Swedish and level 2 in Arabic. Interestingly, he uses level 2 and level 4 but did not produce any phrasal morphology ( level 3). In the control group, nine out of ten children have a high level in one or both languages; four children have a high level in both languages, and five children have a high level in one of the languages. ND:10 Adil belongs to the control group but has a low level in both languages.
280 G. Ha˚kansson et al. Table 11. Comparison of PT levels in Swedish and Arabic Children with language impairment
Level in Swedish
Level in Arabic
Children without language impairment
Level in Swedish
Level in Arabic
LI:1 Zuhair LI:2 Anwar LI:3 Taha LI:4 Mahmoud LI:5 Hanin LI:6 Mohsen LI:7 Abir LI:8 Suha LI:9 Abdul LI:10 Manal
1 1 1 1 2 2 3 2 4 1
1 2 1 2 2 2 2 2 2 1
ND1: Hoda ND:2 Nadja ND:3 Sannah ND:4 Yasin ND:5 Zahla ND:6 Karim ND:7 Mansour ND:8 Azra ND:9 Inaaz ND:10 Adil
5 5 5 5 1 4 4 2 1 2
2 4 3 2 3 3 3 3 3 2
8.8. Additional factors When the age factor is considered, our data on Swedish-Arabic children show a large variability; see Table 12. Two of the children, LI:6 Mohsen and ND:2 Nadja, were exposed to Swedish before the age of two years. They reach level 2 (LI:6 Mohsen) and level 4 (ND:2 Nadja) respectively. Of the children with language impairment that only reach level one, three had their first exposure to Swedish before the age of three, the other two Table 12. Time of exposure to Swedish and age of onset relative to PT level in Swedish Children LI
Time of Age Level in Children exposure of Swedish ND onset
Time of Age of Level in exposure onset Swedish
LI:5 LI:6 LI:9 LI:7 LI:8 LI:1 LI:2 LI:3
3:10 3:7 3:0 2:10 2:6 2:1 1:11 1:6
2:9 1:11 2:5 3:3 2:2 2:6 2:5 2:5
2 2 4 3 2 1 1 1
3:7 2:11 2:10 2:8 1:11 1:6
2:1 2:4 1:11 3:11 3:6 4:2
5 4 5 5 4 5
0:10 0:7
3:0 5:7
1 1
1:2 1:0 0:6 0:6
3:0 2:4 4:7 3:5
2 2 1 1
Hanin Mohsen Abdul Abir Suha Zuhair Anwar Taha
LI:10 Manal LI:4 Mahmoud
ND:1 ND:6 ND:2 ND:4 ND:7 ND:3
Hoda Karim Nadja Yasin Mansour Sannah
ND:10 Adil ND:8 Azra ND:5 Zahla ND:9 Inaaz
Key: Below the empty space are data from the children exposed to Swedish for less that 1:6 years.
Measuring language development 281 after three years of age (LI:4 Mahmoud is 5:7). Of the children from the control group, two reach level 5 after an age of onset of 3:11 years and more. This does not follow the predictions from de Houwer (1995) that two years of age should be a critical age for differentiating between bilingual second-language acquisition and second-language acquisition. When time of exposure is considered, the results show more consistency. After 1:6 years in preschool, the children in the control group are producing structures at levels 4 or 5 in the Swedish test. In the group of children with language impairment, however, only one child, LI:9 Abdul, manages to score at level 4 and this is after an exposure to Swedish of 3:0 years. These results might also be influenced by other factors. Table 13 shows the distribution of age and educational level of the parents of children exposed to Swedish at their preschools for 1:6 years and more. As can be seen, the differences between the two groups regarding both time in Swedish preschool and the mean for parental level of eduction are small. The mean age for children in the control group is slightly higher, 0.36 years (or 4.32 months), than in children with language impairment.
9. Discussion 9.1. The hypotheses Our results show that for eight out of ten children with the diagnosis of language impairment and for nine out of ten children with normal language development, the predictions of both our hypotheses were confirmed. The first hypothesis stated that the grammatical structures would develop in an implicational order, and the second that the children with language impairment would have a low level in both languages, while Table 13. Age and educational level of parents relative to PT level in Swedish for children exposed to Swedish >1.6 years in preschool
Children with language impairment (n=8) Children in the control group (n=6)
Mean age
Mean time in preschool
Educational level of parents
Level in Swedish according to PT
5.03 span 3.11–6:7
2.65
2.25
2
5.39 span 4:5–6:7
2.57
2.33
4.67
282 G. Ha˚kansson et al. the children without language impairment would have a high level in at least one language. The aim of the study was to find a way to measure the children’s two languages. It showed that our model is to a great extent able to differentiate between the children with language impairment and the children in the control group.
9.2. Methodological considerations There are certain conditions that must be fulfilled in order for a structure to be selected as a test item. First, there must be obligatory contexts for the structure in question, otherwise there is no way to decide whether it is left out for reasons of lack of processing capacities or for other reasons, such as a restricted lexicon and situational factors. Second, only structures with an opposition can be used. For Swedish, structures that meet these conditions have been identified from all five levels of the processability hierachy (Pienemann and Ha˚kansson 1999). For Arabic, however, we have as yet found structures for only three levels, word level, level of lexical morphology, and level of phrasal morphology. It is important to remember that the reason for the inability of some of the children to produce the target structure might not be linguistic. Other factors, such as elicitation problems and only a few items on every level, may be the reason. The children in this study are surrounded by a large number of languages other than Arabic and have very limited contact with Swedish outside the preschool. By using a control group for comparison, the test results of children with language impairment can be compared to the test results of children from a similar linguistic environment. This is a way of trying to obtain base-line data of bilingual SwedishArabic children’s language development in both languages. It is, however, almost impossible to get perfect matches on all variables between the groups. The fact that there are six boys with language impairment but only four boys in the control group may have affected the results, since the proportion of boys with language impairment is 2:1 compared to girls (Bishop 1997). Two girls in the control group produced structures at the highest level in both Swedish and Arabic, but none of the boys. On the other hand, two boys with language impairment, who did not correspond to the predictions, showed a high level in one of their languages, but none of the girls in this group showed this type of performance.
Measuring language development 283 9.3. Clinical implications Considering the earlier mentioned lack of language tests, our results show that this model of language development has a potential of being used as a yardstick for different mother tongues to measure how far the children have proceeded on their route toward the target language. Although the children in this study live in ethnically diverse areas, the results in our study indicate that the acquisition of Swedish among the children in the control group was faster than among the children with language impairment. According to Bishop (1997; 44), grammatical development is relatively insensitive to both the quality and the quantity of language input from the environment once a minimum limit of language input has been passed. The children with language impairment show an interesting profile, with a long time of exposure and still a low level of grammatical proficiency in Swedish. This could have clinical significance for the assessment of language impairment if the results are confirmed in a larger study. Our analysis of age of onset gives weak support to the hypothesis of age of onset as a predictor in the control group. All the children in the control group with a low level of Swedish, ND:5 Zahla, ND:9 Inaaz, and ND:10 Adil, have a late age of onset. However, they also have a shorter time of exposure than the other children. There are also two children with a high level of Swedish, ND:4 Yasin and ND:7 Mansour, and a late age of onset. For the children with language impairment it is difficult to see a relationship between first exposure, length of exposure, and level of Swedish. Almost all of these children have a low level of Swedish irrespective of age of onset. Summarizing, neither age of onset nor length of exposure explains the low level of development of Swedish in the children with language impairment. In an earlier study by Salameh et al. (2002), parental report on the child’s problems with language development also proved to have clinical significance for language impairment in bilingual children.
9.4. Theoretical implications Earlier studies of children with language impairment have a bias toward English-speaking children. The results have shown that these children have major difficulties in the area of morphology, especially in the verbal morphology. There are now an increasing number of studies in other languages (Leonard 1998). Different accounts have been suggested in order to explain the underlying factors behind the impairment.
284 G. Ha˚kansson et al. Most children in our study had problems with phrasal but not lexical morphology in both their languages. This adds a new dimension to the surface hypothesis (e.g. Leonard, Bortolini et al. 1992; Leonard, McGregor et al. 1992), according to which children with language impairment have difficulties in perceiving and producing elements of low phonetic substance, and the impairment is due to perceptual limitations. It is thus not only perceptual limitations that are involved but also limitations in processing of the grammatical levels. The results from the present study show the importance of investigating the processes underlying surface structures in cross-linguistic comparisons. The analysis revealed that what is traditionally thought of as ‘‘morphology’’ can be different processes on lexical, phrasal, and interphrasal levels. The children in our study do not exhibit particular difficulties when expressing lexical morphology. Seven of the children with language impairment and all ten control children produced lexical morphology. Similar to children with language impairment in pro-drop languages other than Arabic, such as Italian and Hebrew (Leonard 1998), the children in our study produced verb suffixes expressing person. The dividing line between the children with language impairment and the control group is the phrasal morphology, which is expressed by only one of the ten children with language impairment but by nine of the ten control children. It is interesting to note that phrasal morphology has been shown to be the critical factor in research on verb markers of tense in children with language impairment. For example, Hansson (1997) found that Swedish children have more problems with compound tense (phrasal morphology) than with simple tense. Likewise, Paradis and Crago (2000) found the same tendency in French children with language impairment.
10. Concluding remarks Our results suggest that the developmental perspective needs to be considered in studies of language impairment. The study also shows that such factors as age of onset and exposure do not seem to have any relevance for the children with language impairment. By using implicational scaling to capture the variation, we found a quantitative difference between the children with language impairment and the control group. All the children can be placed along the same continuum of processing complexity. This shows that it is fruitful to use a developmental perspective and study children with language impairment as individuals on different levels,
Measuring language development 285 instead of regarding them as a homogeneous population with a common deficit. Received 3 April 2001 Revised version received 22 January 2002
Lund University Lund University Hospital
Notes * This study was supported by the Bank of Sweden Tercentenary Foundation, grant no. 1998-0188:01, and the Foundation of Queen Silvia’s Jubilee Fund for Research on Children and Handicaps. We would like to thank two anonymous reviewers for their very helpful comments. We also would like to thank Fethi Mansouri for discussions on Arabic linguistics, Ghassan Salameh for help with the assessment of test results in Arabic, and Bo Gullberg for help with statistics. Correspondence address: Professor Gisela Ha˚kansson, Department of Linguistics, Lund University, Helgonabacken 12, 5-223 62 Lund, Sweden. E-mail:
[email protected]. 1. One referee pointed out that it is a methodological problem for our study in that we only show the first three levels. We are aware of this problem and regret that we have not been able to find structures representing levels 4 and 5 that are used in the oral production of the children. Earlier studies of L2 acquisition of Arabic (Mansouri 1995, 1999, 2000; Lykke-Nielsen 1996) have dealt with adults learning Modern Standard Arabic at university level. These adult learners used full subjects ( level 4) as well as subordinate clauses ( level 5).
References Andersson, Anders-Bo¨rje (1992). Second Language Learners’ Aquisition of Grammatical Gender in Swedish. Gothenburg Monographs in Linguistics 10. Gothenburg: University of Gothenburg. Bishop, Dorothy (1997). Uncommon Understanding. Development and Disorders of Language Comprehension in Children. Hove: Psychology Press. Bolander, Maria (1988). Is there any order? On word order in Swedish learner language. Journal of Multilingual and Multicultural Development 9, 97–113. Bolotin, Naomi (1995). Arabic and parametric VSO agreement. In Perspectives on Arabic Linguistics VII, Mushira Eid (ed.), 7–27. Amsterdam: Benjamins. Bortolini, Umberta; and Leonard, Laurence (1996). Phonology and grammatical morphology in specific language impairment: accounting for individual variation in English and Italian. Applied Psycholinguistics 17, 85–104. Clahsen, Harald (1988). Normale und gesto¨rte Kindersprache. Amsterdam and Philadelphia: Benjamins. —(1989). The grammatical characterization of developmental dysphasia. Linguistics 27, 897–920. —(1992). Linguistic Perspectives on Specific Language Impairment. Theorie des Lexicons. Arbeiten des Sonderforschungs-bereichs 282. Du¨sseldorf: University of Du¨sseldorf.
286 G. Ha˚kansson et al. —; Bartke, Susan; and Go¨llner, Sandra (1996). Formal features in impaired grammars. Essex Research Reports in Linguistics 14, 42–79. Crystal, David; Fletcher, Paul; and Garman, Michael (1976). The Grammatical Analysis of Language Disability. London: Cole and Whurr. (Second edition 1988.) Damico, Jack, S.; and Hamayan, Else V. (1990). Implementing assessment in the real world. In Limiting Bias in the Assessment of Bilingual Students, Else V. Hamayan and Jack S. Damico, (eds.), 303–318. Austin, TX: Pro-ed. de Houwer, Annick (1995). Bilingual language acquisition. In The Handbook of Child Language, Paul Fletcher and Brian MacWhinney (eds.), 219–250. Oxford: Blackwell. de Montfort Supple, Marie (1996). Prologue: beyond bilingualism. Topics in Language Disorders 16(4), 1–8. Dromi, Esther; Leonard, Laurence; and Shteiman, Michael (1993). The grammatical morphology of Hebrew-speaking children with specific language impairment: some competing hypotheses. Journal of Speech and Hearing Research 36, 760–771. Eid, Mushira (1990). Arabic linguistics: the current scene. In Perspectives on Arabic Linguistics I, Mushira Eid (ed.), 3–37. Amsterdam: Benjamins. Ferguson, Charles (1959). Diglossia. Word 15, 325–340. Fischer, Wolfdietrich; and Jastrow, Otto (eds.) (1980). Handbuch der Arabischen Dialekte. Porta Linguarum Orientalum, Neue Serie XVI. Wiesbaden: Harrassowitz. Gerken, Liza (1991). The metrical basis for children’s subjectless sentences. Journal of Memory and Language 30, 431–451. Glahn, Esther; Ha˚kansson, Gisela; Hammarberg, Bjo¨rn; Holmen, Anne; Hvenekilde, Anne; and Lund, Karen (2001). Processability in Scandinavian second language acquisition. Studies of Second Language Acquisition 23, 389–416. Gopnik, Myrna (1990). Feature-blind grammar and dysphasia. Nature 34, 715. Grimm, Hannelore; and Weinert, Sabine (1990). Is the syntax development of dysphasic children deviant and why? New findings to an old question. Journal of Speech and Hearing Research 33, 220–228. Ha˚kansson, Gisela (2001). Tense morphology and verb-second in Swedish L1 children, L2 children and children with SLI. Bilingualism, Language and Cognition 4, 85–99. —; and Hansson, Kristina (2000). Comprehension and production of relative clauses: a comparison between Swedish impaired and unimpaired children. Journal of Child Language 27, 313–333. —; and Nettelbladt, Ulrika (1993). Developmental sequences in L1 (normal and impaired) and L2 acquisition of Swedish. International Journal of Applied Linguistics 3, 131–157. —; and Nettelbladt, Ulrika (1996). Similarities between SLI and L2 children. Evidence from the acquisition of Swedish word order. In Children’s Language, vol. 9, John Gilbert and Carolyn Johnson (eds.), 135–151. Mahwah, NJ: Erlbaum. Hansson, Kristina (1997). Patterns of verb usage in Swedish children with SLI: an application of recent theories. First Language 17, 195–217. —(1998). Specific Language Impairment in Swedish. Grammar and Interaction. Studies in Logopedics and Phoniatrics 2. Lund: Department of Logopedics and Phoniatrics, Lund University. —; and Nettelbladt, Ulrika (1995). Grammatical characteristics of Swedish children with SLI. Journal of Speech, Language, and Hearing Research 38, 589–598. Hellquist, Britt (1989). Nya SIT — Spra˚kligt Impressivt Test fo¨r Barn [Language Comprehension Test for Children]. Lo¨ddeko¨pinge: Pedagogisk Design. Holes, Clive (1995). Modern Arabic. New York: Longman. Holmberg, Eva; and Bergstro¨m, Anneli (1996). Oralmotoriskt test fo¨r barn (ORIS) [Oral motor test for children]. Lo¨ddeko¨pinge: Pedagogisk Design.
Measuring language development 287 —; and Stenkvist, Harrieth (1983). Nva Lundamaterialet. Kartla¨ggning och bedo¨mning av barns spra˚kliga fo¨rma˚ga [ The Lund Test of Phonology and Grammar. Description and assessment of children’s linguistic abilities]. Malmo¨: Utbildningsproduktion AB. Hyltenstam, Kenneth (1977). Implicational patterns in interlanguage syntax variation. Language Learning 27, 383–411. Leissner, Inga; Nilsson, Birgitta; Nystro¨m, Gisela; and Wastesson, Birgitta (1962). Leiters Klosstest [Leiter International Performance Scale]. Stockholm: Psykologifo¨rlaget. Leonard, Laurence (1989). Language learnability and specific language impairment in children. Applied Psycholinguistics 10, 179–202. —(1998). Children with Specific Language Impairment. Cambridge, MA: MIT Press. —; Bortolini, Umberta; Caselli, Maria Cristina; McGregor, Karla; and Sabbadini, Letizia (1992). Morphological deficits in children with specific language impairment: the status of features in the underlying grammar. Language Acquisition 2, 151–179. —; McGregor, Karla; and Allen, George (1992), Grammatical morphology and speech perception in children with specific language impairment. Journal of Speech and Hearing Research 35, 1076–1085. Levelt, Willem (1989). Speaking: From Intention to Articulation. Cambridge, MA: MIT Press. Long, Michael (1993). Second language acquisition as a function of age: research findings and methodological issues. In Progression and Regression in Language, Kenneth ˚ ke Viberg (eds.), 289–308. Cambridge: Cambridge University Press. Hyltenstam and A Long, Steven (1994). Language and bilingual-bicultural children. In An Introduction to Children with Language Disorders, Vicki A. Reed (ed.). New York: McMillan. Lykke-Nielsen, Helle (1996). Acquisition order in Arabic (L2): the multidimensional aproach. Odense Working Papers in Language and Communication 11. MacLaughlin, Barry (1984). Second Language Acquisition in Childhood, vol. 1, 2nd ed. Hillsdale, NJ: Erlbaum. Mansouri, Fethi (1995). The acquisition of subject–verb agreement in Arabic as a second language. Australian Review of Applied Linguistics 16, 64–85. —(1999). The Acquisition of Arabic as a Second Language: From Theory to Practice. ASLA Australian Studies in Language Acquisition 7. Sydney: LARC University of Western Sydney. —(2000). Grammatical Markedness and Information Processing in the Acquisition of Arabic as a Second Language. Lincon Studies in Language Acquisition 2. Munich: Lincon. Meisel, Ju¨rgen; Clahsen, Harald; and Pienemann, Manfred (1981). On determining developmental stages in natural second language acquisition. Studies in Second Language Acquisition 3, 109–135. Mohammad, Mohammad A. (1990). The problem of subject–verb agreement in Arabic: towards a solution. In Perspectives on Arabic Linguistics I, Mushira Eid (ed.), 95–125. Amsterdam: Benjamins. Mo¨hring, Anja (2001). The acquisition of French by German pre-school children. An empirical investigation of gender assignment and gender agreement. In EUROSLA Yearbook 1, 171–193. Amsterdam: Benjamins. Nettelbladt, Ulrika (1983). Developmental Studies of Dysphonology in Children. Lund: Liber La¨romedel. —(1998). Current theories of specifc language impairment (SLI ) in children. Logopedics Phoniatrics Vocology 23, 97–105. Omar, Margaret (1973). The Acquisition of Egyptian Arabic as a Native Language. The Hague: Mouton.
288 G. Ha˚kansson et al. Paradis, Joanne; and Crago, Martha (2000). Tense and temporality: a comparison between children learning a second language and children with SLI. Journal of Speech, Language, and Hearing Research 43, 834–847. Pienemann, Manfred (1998). Language Processing and Second Language Development: Processability Theory. Amsterdam: Benjamins. —; and Ha˚kansson, Gisela (1999). A unified approach towards the development of Swedish as L2: a processability account. Studies of Second Language Acquisition 21, 383–420. —; Johnston, Malcolm; and Brindley, Geoff (1988). Constructing an acquisition-based procedure for second language assessment. Studies in Second Language Acquisition 10, 217–243. Platzack, Christer (1992). Functional categories and Early Swedish. In The Acquisition of Verb Placement, Ju¨rgen Meisel (ed.), 63–82. Dordrecht: Kluwer Academic. —(1996). The initial hypothesis of syntax. A minimalist perspective on language acquisition and attrition. In Generative Perspectives on Language Acquisitions, Harald Clahsen (ed.), Amsterdam: Benjamins. Rice, Mable; and Wexler, Ken (1996). Towards tense as a clinical marker of specific language impairment in English-speaking children. Journal of Speech and Hearing Research 39, 1239–1257. Romaine, Suzanne (1989). Bilingualism. Oxford: Blackwell. Sahle´n, Birgitta; and Nettelbladt, Ulrika (1995). How specific is specific language impairment? On problems of delineation and classification. Scandinavian Journal of Logopedics and Phoniatrics 20, 51–60. Salameh, Eva-Kristina (2000). Utva¨rdering av metodik vid testning av barn med svenska som andraspra˚k [Assessement of methods for testing children with Swedish as a second language]. www.ldc.lu.se/logopedi/department —; Ha˚kansson, Gisela; and Nettelbladt, Ulrika (1996). The acquisition of Swedish as second language in a group of Arabic-speaking pre-school children: word order patterns and phrasal morphology. Logopedics Phoniatrics Vocology 21, 163–170. —; Nettelbladt, Ulrika; Ha˚kansson, Gisela; and Gullberg, Bo (2002). Language impairment in Swedish bilingual children: a comparison between bilingual and monolingual children in Malmo¨. Acta Pediatrica 91, 229–234. Santelmann, Lynn (1995). The acquisition of verb second grammar in child Swedish: continuity of universal grammar in wh-questions, topicalization and verb raising. Unpublished dissertation, Cornell University. Schlyter, Susanne (1993). The weaker language in bilingual Swedish-French children. In ˚ ke Viberg (eds.), Progression and Regression in Language, Kenneth Hyltenstam and A 289–308. Cambridge: Cambridge University Press. Schumann, John (1997). The neurobiology of affect in language. Language Learning 48. Volterra, Virginia; and Taeschner, Traute (1978). The acquisition and development of language by bilingual children. Journal of Child Language 5, 311–326. Westernoff, Fern (1991). The assessment of communication disorders in second language learners. Journal of Speech Language Pathology Audiology 15(4), 73–79. Wode, Henning (1981). Learning a Second Language. An Integrated View of Language Acquisition. Tu¨bingen: Narr.
Copyright of Linguistics is the property of De Gruyter and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
