Turkish in contact with German: Language maintenance and loss among immigrant children in Berlin...

33
Turkish in contact with German: Language maintenance and loss among immigrant children in Berlin (West)' CAROL W. PFAFF lntroduction The phenomenon of massive migration of workers frorn Turkey to the Federal Republic of Germany, the Netherlands, and other northwestern European countries has brought about a situation of particular interest to sociolinguistics and psycholinguists concerned with language mainte- nance, language loss, and language-contact phenomena. Here the in- terplay of social factors with linguistic universals and language particulars is brought into sharp focus because the languages in contact are geneti- cally unrelated and typologically very different from each other. In this Paper, I present some results of two studies of second- and third-genera- tion preschool and primary school children of Turkish rnigrants in West Berlin. I look at structural linguistic effects in the grammatical subsystems which express nominal reference (lexicon, case and number marking, modification, anaphora). These differ in their inherent complexity and in the extent to which their expression differs structurally from the contact language, German. Before turning to the studies themselves, I will first set the stage with some brief comments on the social and dernographic background of migrants in Berlin and on the educational policies which establish the framework in which the children's language acquisition, language contact, and language use rnust be placed. Second, I will briefly describe the studies and the particular subsamples of preschool and primary school children whose language production is considered. Third, I sketch the typological features of Turkish relevant to the investigation of nominal reference and our hypotheses about their devel- opment in TurkishJGerman bilingual children.' Sections 4 through 8 present the results and section 9 gives a Summary and perspectives on the issues of language development in this contact setting. 0 16s-2.5 i 6/91 10090-0097 $2.00 Int'l. J. SOC. Lang 90 (1991). pp. 97-129 . 0 Wder de Gruyrer

Transcript of Turkish in contact with German: Language maintenance and loss among immigrant children in Berlin...

Turkish in contact with German: Language maintenance and loss among

immigrant children in Berlin (West)'

CAROL W. PFAFF

lntroduction

The phenomenon of massive migration of workers frorn Turkey to the Federal Republic of Germany, the Netherlands, and other northwestern European countries has brought about a situation of particular interest to sociolinguistics and psycholinguists concerned with language mainte- nance, language loss, and language-contact phenomena. Here the in- terplay of social factors with linguistic universals and language particulars is brought into sharp focus because the languages in contact are geneti- cally unrelated and typologically very different from each other. In this Paper, I present some results of two studies of second- and third-genera- tion preschool and primary school children of Turkish rnigrants in West Berlin. I look at structural linguistic effects in the grammatical subsystems which express nominal reference (lexicon, case and number marking, modification, anaphora). These differ in their inherent complexity and in the extent to which their expression differs structurally from the contact language, German.

Before turning to the studies themselves, I will first set the stage with some brief comments on the social and dernographic background of migrants in Berlin and on the educational policies which establish the framework in which the children's language acquisition, language contact, and language use rnust be placed.

Second, I will briefly describe the studies and the particular subsamples of preschool and primary school children whose language production is considered.

Third, I sketch the typological features of Turkish relevant to the investigation of nominal reference and our hypotheses about their devel- opment in TurkishJGerman bilingual children.' Sections 4 through 8 present the results and section 9 gives a Summary and perspectives on the issues of language development in this contact setting.

0 16s-2.5 i 6/91 10090-0097 $2.00 Int'l. J. SOC. Lang 90 (1991). pp. 97-129 . 0 Wder de Gruyrer

1. Sociolinguistic background of rnaintenance and loss of Turkish in Berlin (West)

Among the most important interrelated social factors which provide the framework in which maintenance o r loss of a minority language proceeds are population density and opportunity for contact with speakers of majority and minority languages in contact. The situation of rnigrants from Turkey in Berlin (West) is quite complex. A more extensive treat- ment of the historical development and sociopolitical situation can be found in Kardam et al. (1988).

1.1. Dernographic background

The 1988 census figures for Berlin (West) show that 1 1 percent of the population do not have German citizenship, with Turks composing the largest Segment of this population, about 50 percent. The second largest group, migrants from Yugoslavia, constitutes only about 13 percent of the foreign population.

The migrants classified here as Turkish on the basis of their nationality are, however, neither linguistically nor socially homogeneous and include rnembers of ethnic and linguistic minorities such as Circassians and Kurds, the latter estimated to compose as much as one-fifth to one-third of the Turkish population in the Federal Republic of Germany (Blashke and Ammann 1988: 92). While many such ethnic-minority migrants also speak Turkish, their dialects may diKer more or less from standard Turkish as a result of speaking (or having spoken) an ethnic language unrelated to Turkish as well. a possibility which rnust be taken into account when assessing the linguistic changes in Turkish in the current contact Setting. Indeed, this problem is general, as many of the ethnically Turkish migrants speak varieties which depart from the standard varieties described in the major grammars.

Nor do the migrants from Turkey form a socioeconomically homoge- neous group: while the majority are former peasant and working-class migrants motivated primarily by economic concerns, many of whom originally intended to return to Turkey with a better socioeconomic standing, there are also numerous politically motivated migrants, includ- ing highly educated intellectuals who, at least for the present, have no plans to return.

The foreign population of West Berlin is distributed unevenly through- out the city. While the overall percentage of migrants from Turkey in West Berlin is 5.5 percent, the proportions in the districts of Wedding

Turkish in contact n-ith Germatz 99

and Kreuzerg are considerably higher, 12.8 percent and 18.8 percent, respectively. As a reflection of the demographic distribution and the relatively high birthrate among foreigners, as opposed to a relatively low birthrate among Germans, the proportions are even higher for children and adolescents, as shown in Table 1, which focuses on the Turkish population in these two districts, and in two neighborhoods, Gesundbrun- nen in Wedding and Mariannenplatz in Kreuzberg, in which most of the data in the studies discussed here were collected.

Turkish children comprise about 20 percent of the total age cohort in Gesundbrunnen and about 50 percent in Mariannenplatz.

The sociolinguistic consequence is that the opportunities for neighbor- hood contact with other Turkish children, particularly in areas such as Mariannenplatz in Kreuzberg, are likely to be as great as or greater than for contact with native speakers of German, a factor which influences

Table 1. Turkish popularion by age in Berlh ( West), Kreuzberg. und Wedding und in selecred neighborlioodr' (source: Statistisches Landesamt, 1988)

Turks % Turks of % Turks of n foreigners total population

Berlin (West) all ages 120,351 45 6 under 6 14,679 62 13 6-15 21,131 60 14

Wedding all ages 20.624 61 13 under 6 2.540 74 26 6-15 3,507 72 30

Gesundbrunnen all ages 6.492 61 28 under 6 448 74 19 6-15 639 72 22

Kreuzberg all ages 27.732 66 19 under 6 3,478 79 3 3 6-15 5,049 77 39

Mariannenplatz all ages 3,916 66 13 under 6 1.018 79 48 6-15 1.450 77 53

a. Data collection for the projects discussed was conducted at preschools and afternoon day-care Centers in these two neighborhoods. i h e number of Turks in these neighbor- hoods is estirnated frorn the exact figures for non-Gennan residents.

the extent of firsr-language maintenance which we find characteristic of children from such neighborhoods (See Pfaff 1988a for a discussion of self-reported language-use behavior among 5- 12-year-olds).

1.2. School policies for linguisric-minoriry children

The tendency toward language maintenance rather than shift is further supported by a school policy which, despite the official goal of integration, nonetheless allows for Ausländerregelklassen (regular foreigner classes) which are made up entirely of foreign pupils, often mainly Turkish speakers. The rationale behind such classes is that German pupils should not be in the minority in their classrooms. Thus the proportion of foreign children in Regelklassen (regular classes) attended by Germans is limited to 30 percent or 50 percent (if the foreign children have no significant difficulties in following instruction in German) (Senator für Schulwesen, Berufsbildung und Sport 198711 988). In the school year 19871 1988, 5,8 18 foreign children attended such classes without Gennan Peers, particularly in the districts of Wedding and Kreuzberg, where the proportions reached one-third to one-half of the foreign pnmary school pupils and one-fourth to one-third of the foreign secondary pupils.

Although the majonty of foreign children in such Ausländerregelklassen are Turkish, and some of their teachers are Turkish (see Ucar 1986), the regular curriculum is followed and the language of instruction is German. Mother-tongue education has not generally been provided in Berlin, in contrast to the policy in other northwestern European countries such as Sweden and the Netherlands, and in contrast to the Situation in other West Gennan Bundesländer outside of Berlin such as Bavaria, which, from the beginning, adopted a policy oriented to the eventual return of the migrant populations to their home countries. The concentration on German instruction in Berlin reflects the policy of integration of foreign pupils. Instruction in Turkish was, until recently, limited to Vorbereitrotgs- klassen (preparatory classes) for older children and adolescents who were entering the Berlin school System with no knowledge of German, or for those younzer children whose German was inadequate to follow regular instruction in German. Turkish teachers were pennitted to provide assis- tance in Turkish.

It is only in the last decade that instruction in Turkish has been introduced into the regular curriculum, first at the secondary level (Tur- kish instead of a foreign language) and later at the primary level in initial \ ,

reading and writing instruction as well as in school subjects. The offerings at both secondary and primary levels remain rather limited at present. .

Turkish in cortract ic.itlt Gertltan 101

Classes in Turkish instead of a foreign language. introduced experimen- tally in 1979, were attended by 2,116 pupils (1 6 percent of the Turkish pupils in grades 5- 10) in the school year 1987/1988.

Linguistic analysis of the Turkish language skills of pupils in such classes (Steinmüller 1987; Cakir and Meyer 1988) reveal language attrition in vocabulary, orthography, and some areas of Syntax in comparison with the norms of standard Turkish. More detailed investigations of the oral and written production and comprehension of these pupils would provide valuable insight into the processes of language attntion in individ- uals and in the Turkish migrant community.

Gaps in pupils' knowledge of Turkish has provided one argument for the introduction of early mother-tongue instruction (Thomas 1987: 50-5 1; Steinmüller 1987: 278). in addition to the rationale based on the views of Cummins (1984) and Skutnabb-Kangas (1981) that development in the first language facilitates second-language acquisition.

Initial bilingual education in primary schools, introduced in 1984/1985, is still in the experimental Stages and is limited to one or two classes in grades 1-4 at three primary schools. The programs were scheduled for evaluation at the end of the school year 1988/1989. At present, they are attended pnmarily by those children whose Gerrnan is most limited when they enter school and essentially replace the previous preparatory classes (see Keskin 1988; Nehr et al. 1988).

As a result of the demographic distribution of migrants from Turkey and of the educational policies outlined above, it seems that maintenance of spoken Turkish is fostered in those areas with high concentrations of Turks, but that, in the absence of formal mother-tongue instruction. the language is particularly susceptible to the processes of language change. both those which derive from internal linguistic pressures (loss of marked forms and structures) and those which result from incorporation of elements from the contact language, German. In the rest of this Paper, I discuss some findings of two studies of Turkish/German bilingual children in Berlin, focusing on aspects of nominal reference.

2. Studies of TurkishIGerman bilingual children in Berlin

Since 1978, we have been investigating aspects of migrant children's language in Berlin, considering both the acquisition of German as a second language by children of various linguistic backgrounds and. more recently, the development of Turkish as a first language as well. An overview of the two studies discussed here is given in Table 2.

The EKMAUS study is a cross-sectional investigation of the speech of

102. C. W.' P f j -

Table 2. Selecled sludies o/ bilingual children in Berlin

Data collection lnformants Setting

Methodology

EKMAUS project' Preschool and afterschol day-care centers

. 1983-1986

KITA project' Preschool and day-care centers 1987 - longitudinal study

5-12-year-old Turkish/German bilinguals: (A, B, C)b

monolinguals: Turkish (D) German (E)

I -6-year-old TurkishIGerman bilinguals: TT. DD. TD. TT = both parents

Turkish DD = both parents

German T D = one parent

Turkish, one parent German

Semistructured experimental interviews in Turkish and German

Recordings of semistructured adult-child interactions in Turkish and German

a. This project, entitled 'Linguistische und kognitive ~ntwicklun~: die Beziehung zwischen Erst- und Zweitspracherwerb'. was one of several funded by the Freie Universität Berlin under the rubnc EKMAUS (Entwicklung von Konzepten und Materialien für die Förderung ausländischer Kinder und Jugendlichen im schulischen und außerschulischen Bereich).

b. Group A = in Germany from birth or early age, little contact with native German peers. Group B=in Grnnany from birth or early age, considerable contact with native Gerrnan peers. Group C=entered Germany after early childhood and two or rnore years of school in Turkey. Group D = monolingual control group, Turkish children in Ankara. Group E=rnonolingual control group, Gerrnan children in Berlin.

C. This project, entitled 'Natürlicher bilingualer Sprachenverb von KITA-Kindern: Krip- penalter bis ersie Grundschuljahre', has been funded since its inception in May 1987 by the Deutsche Forschungsgerneinschaft under its 'Schwerpunkt: Spracherwerb'.

around 80 TurkishIGerman bilingual children from Wedding and Kreuz- berg between the ages of 5 and 12 with varying degrees of contact with German in their neighborhoods and at school and difTering in the age at which contact with the second language began (group A: little contact; B: more contact; C: immigrated to Berlin only after initial socialization in Turkey); monolingual control groups were also investigated (groups D: Turkish and E: Germanj, as indicated in Table 2. This study was experimental, using games with toys, pictures, and narratives based on verbal stimuli to elicit oral production, comprehension, and metalinguistic

Turkish in contact with German 103

judgments of various aspects of nominal reference (See Pfaff et al. 1985 for a full discussion of the methodology).

The KITA study, begun in 1987 and currently in Progress, is longitudi- nal and investigates the language of around 20 Turkish and German children between the ages of one and seven who attend a bilingual day- care Center (Kita) in Kreuzberg. Children from families where both parents are Turkish (Ti) compose about 80 percent of the Kita popula- tion; the rest of the children are Gennan (DD) or children of mixed rnarriages (TD) or of other nationalities. The staff is composed of equal numbers of native speakers of Turkish and Gennan, who follow the principles of equal nghts for both lariguages. As a rule, each caretaker speaks her own mother tongue to all children; thus children are exposed to native adult German and Turkish from a very early age.' Despite this early exposure to both languages, we find that the children are clearly dominant in one or the other language, so we speak of first- and early second-language acquisition rather than of sirnultaneous language acquisition (PfaK et al. 1987, 1988). Our data-collection activities include participant observation in the regular activities in the Kita and recordings averaging half an hour in length rnade with each child in each language once a rnonth (Pfaff et al. 1987; Pfaff and Savas 1988).

2.1. Sociolinguistic background of selected children

In this Paper, I will focus on individual children drawn frorn both the KITA and EKMAUS studies, taking a look at sorne characteristics of the varieties of Turkish closer than is possible in the comparison of group data.

Two ages will be examined: four- to five-year-olds frorn the KITA project and eight- to nine-year-olds frorn the EKMAUS project. Sorne background information of these children is sumrnarized in Tables 3 and 4. The KITA children, four to five years old during the period of investiga-

Table 3. Sekc~ed four- fo Jive-yeor-olds from KITA projecf

Childlcode Date of Enrolled in Language Vacation Preferred birth KITA since at home in language

age Turkey at KITA -

TT 1 6 1 13.7.83 I; 2 TK no TK TT S m 13.3.83 2; 6 DT/(TK) yes DT T D 2 1 20.9.83 2; 2 DT Y es DT

Table 4. Sclecretl'ei.q/~r- ro iriuc-!~cnr-ol(l.v fronr EKMAUS prqccr - -

Child/codc8 Sex/ Birth- Languuge in juniily Language Grade Classb Teacher Lireracy age place with wiih wiih type reads, writes 7

parcnfs siblings friends

A S 2 I I 1 9 ßerlin TK TK TK I st A DT/TK DT DT ß S6 04 ßerlin DT/(TK) TK DT/TK 2nd A DTflK . DT/(TK) DT C 5801 m/8 Turkcy TK/DT TK/DT TK/DT 2nd R DT DT/DK DT/TK D S7 09 m/9 Turkey TK TK TK 3rd - TK TK TK

a. Groups A. B, and C are based on contact with native Germans and age of onset of acquisition of German. SI. S? ... code ihe day-care Centers whcre the interviews were conducted. 01. 02, ... plus sex and age code the individual child. b. Class types: A = AitslÜn~lerregclkIa.tse~~; R = Rcgelklasse~i.

Turkish in conracr wirh German 105

tion under consideration, June 1987-July 1988, represent three different constellations of sociolinguistic factors:

TT 16 f (3; 10-4;7) is a member of a family which lives in close contact with other relatives in Berlin. The prevailing language environment is Turkish.

TT 5 m (4;4-5;2) in contrast, has spent several intennittent penods of time living away from his family in a children's home, where his contact with German children and adults was increased, and, at present, his family has considerable contact with native Germans.

TD 2 f (3;lO-4;8) is the child of a mixed marriage. As is quite typical in such cases, German, the language of communication between her parents, is her dominant language, although her mother may sometimes address her in Turkish. She clearly understands Turkish; however, dunng the period investigated, she strongly preferred to speak German, even in interactions with the Turkish interviewen.

The four eight- to nine-year-old children selected from the EKMAUS project represent the four groups A, B, C, and D as defined above.

A S2 11 f 9 was born in Berlin, lives in an area of Kreuzberg which is predominantly Turkish, attends one of the Ausländerregelklassen with exclusively foreign pupils, and attends an after-school day-care ccnter (Schülerladen), which was attended only by Turkish children from-thc neighborhood, although both Turkish and German caretakers were pre- sent. Not surprisingly, her dominant language is Turkish.

B S6 04 f 8, also born in Berlin, lives in a more mixed neighborhood. She speaks German more than Turkish with her parents, but Turkish with her brothers, who live in Turkey. The entire family intends to remigrate to Turkey in the near future to a house they have already built there. Although she also attends an ~usländerre~elklasse, the after-school day-care Center (Hort) which she attends has a mixed German and foreign population of children and all the caretakers are German. She speaks both Turkish and German but prefers German.

C S8 01 m 8 had immigrated to Germany only recently at the time of his interviews. He had attended school in Turkey and could already read and write Turkish. In Berlin, he and his older sister are making a strong attempt to learn German as quickly as possible. He attends a Regelklasse with predominantly German children, is the only Turkish child in his second grade class, and speaks German at home as well.

D S7 09 rn 9 represents the monolingual control group in Anakara, Turkey.

In sections 3.1-3.5 we outline our hypotheses about the possible effects of the language-contact Situations we have described on five aspects of .

nominal reference in Turkish, and in sections 4-8 we present the results for the seven children selected to represent the groups investigated.

3. Nominal reference in Turkish: a brief Summary and hypotheses for language loss

Turkish is an agglutinative, highly regular SOV language, typologically as well as genetically very different from German. Brief descriptions of its typology and grammar are provided by Underhill(1986) and Kornfilt (1987).

The subsystems of nominal reference are of interest in themselves and because of their typological contrast with the contact language, German. The nominal-reference features we are investigating are listed below; those discussed in the present paper are marked with '+'. Each is de- scribed briefly below together with our hypotheses about structural conse- quences of language attrition.

Subsystems of nominal reference in Turkish + lexical differentiation, vocabulary + case marking + number marking + modification

definiteness word order within NP and with respect to predicate

+ pronominal vs. Zero anaphora

3.1. Lexical differentiation, vocabulary

The lexicon of Turkish is, in general, not cognate with that of German; thus we expect a rather rapid increase in the use of German lexical items in Turkish as contact with German increases and shift to German pro- gresses. However, because of the typological dissimilarities of the two languages, mixing and structural integration of loan words from German into Turkish may be more constrained than in cases of contact between genetically and typologically related languages such as EnglishJSwedish or Spani~hiEnglish.~

3.2. Case marking

Case marking in Turkish is well known for its transparency and regularity. The allomorphs of the inflections follow regular phonological rules of '

Turkish in contact with German 107

vowel harmony, consonant harmony (voicing assimilation), and the inser- tion of a nasal or glide between sequences of vowels which would arise when vowel-initial suffixes are added to stems ending in vowels, as shown in Table 5.

As illustrated by the forms for 'clothespin', 'ball', and 'eye', the case inflections are generally attached to the stem, which is unaltered in form from the uninflected nominative. There is, however, a minimal amount of phonologically conditioned stem alternation for stems ending in /k/, as in inek 'cow' where the /k/ becomes /E/ realized phonetically as a glide or unrealized before suffixes which begin with vowels.

Studies of monolingual Turkish first-language acquisition by Ekmekci and Aksu-Koc and Slobin (1985) have demonstrated that these case markers are acquired easily and, in comparison to languages with less transparent Systems, very early, by the age of two.

As Slobin (1986: 275) has pointed out, the case system has proved to be quite stable historically in the Course of contact of Turkish with various Indo-European and other languages. We expect a similar outcome here: retention of the formal system of case marking.

3.3. Nurnber marking

The plural Suffixe -ler/-lar is an extremely regular marker with alternants conditioned only by vowel harmony and not involving any concomitant stem changes. In Standard Turkish, this marker is used with plural nouns only when the noun phrase is not otherwise marked for plurality, as for example with plural quantifiers such as iki 'two', üc 'three', or cok 'many'; for example, inek 'cow', inekler 'cows', iki inek 'two c o w ~ ' . ~

As Nilsson (1985), following Grönbeck (1936) and Pntsak (1963), has pointed out, plural marking appears to be a relatively recent histoncal

Table 5. Standard Turkish case f o r m

'cow' 'clothespin' 'ball' 'eye'

Nominative 0 Genitive -(n) In4 Accusative -(Y) I Dative -(Y) E Locative - DE Ablative - DEn Instrumental/-(y) IE comitative

inek inegin inegi inege inekte inekten inekle

mandal mandalin mandali mandala mandalda mandaldan mandalla

top göz topun göziin topu gözü topa göze topta gözde toptan gözden topla gözle

I08 C. W. Pfaff

innovation in Turkish, which began in the ninth century under the infiu- ence of contact with Indo-European languages. It is interesting to See whether this category will continue to change in the diaspora with renewed intensive contact with Indo-European, in particular, whether the environ-

' ments in which the plural morpheme is used will spread to include NPs with plural quantifiers, as in the contact language, German.

Turkish has a variety of devices for modification. The simplest device is the use of adjectives which are uninflected and preposed to the nouns they modify, as in kiiciik inek 'small cow'. As this structure is simple, typologically consistent with the basis SOV order of Turkish, and parallel to the L2 in contact, there is no reason to predict erosion or loss here.

A second type of modification in Turkish involves the derivational sufix -11, which forms adjectives or nouns denoting the possession of the object or quality of the noun to which it is suffixed, as in petnbe elbiseli kiz 'the girl with the pink dress'. Ekmekci (1986) does not list any errors . involving this suffix in the preschool monolinguals she investigated. Aksu- Koc and Slobin (1985: 854), however, note that this construction is subject to occasional errors by monolingual Turkish children, who may overextend the use of the suffix to form semantically anomalous combina- tions. We also predict semantic anomalies similar to those noted for monolinguals.

More complex modification, corresponding to relative clauses in Indo- European languages, is formed in Turkish by deverbal clauses which preceed the noun they modify, for example duran inek 'the cow which is standing' for a subject relative, or Ali'nin gördügü inek 'the cow that Ali saw' for a nonsubject relative (See Underhill 1972; Hankamer and Knecht 1976; Kornfilt forthcoming; for discussion). These participial clauses lack complementizers (equivalent to relative pronouns of Indo-European lan- guages) and have gaps in the position of the constitutent that corresponds to the head.

As Slobin (1986) notes, these nontransparent structures are acquired late in monolingual child production and are frequently misinterpreted in comprehension experiments. He notes further that monolinguals fre- quently have recourse to more analytic, more canonical paraphrases with the particles hani and ya, which are used in adult Speech to emphasize shared knowledge in discourse. We predict that bilingual children would be even slower than monolinguals to acquire deverbal modifying struc- tures and would make extensive use of alternative paraphrases.

Turkish in contact with German 109

3.5. Pronominal vs. fl anaphora

Turkish is a 'pro-drop' language that pennits null representation of anaphonc referents both within and across sentence boundanes for refer- ents which are recoverable from discourse. Unlike many of the best- described pro-drop languages, such as Italian and Spanish, in which the phenomenon is restricted to subjects, in Turkish the null referent may have vanous grammatical functions: subject, object, possessor, etc., so that the Syntax of Turkish differs strikingly from that of German. It has been noted in studies of second-language acquisition (White 1985; Phin-

. ney 1987) that native speakers of pro-drop languages may transfer this feature to their interlanguage. Similarly, one might predict that in lan- guage-contact Situations, the features of the second language might be transferred to the native language, although, if Hyams (1986) is correct and pro-drop is the unmarked Parameter setting, then this effect should be less than transfer in the reverse direction. In any case, the effect on Turkish would not be ungrammatical utterances, but utterances with 'extra' pronouns which are inappropriate in their discourse context.

In the following sections, we will briefly esamine the Turkish results for the subsampks of the four- to five-year-olds from the KITA study and the eight- to nine-year-olds from the EKMAUS study to See if there is evidence of language loss in these nominal-reference features.

4. Language rnixing

With regard to mixing lexical elements from one language into discourse in the other language,' we find over all that mixing is very rare in our recordings but that there are both quantitative and qualitative differences associated with language dorninance and the extent of contact with Germans.

The development of language mixing in the KITA children is clearly increasingly in the direction of the prevailin? asymmetric sociolinguistic norms in the bilingual setting here (See Pfaff et al. 1988: 10-17). The tendency is toward a variety of Turkish which incorporates loan elements from German, but toward a variety of German free of Turkish lexical items:

If we look more closely at the extent to which these loan elements are integrated morphosyntactically into the receiving language, shown in

I

.\

Table 6, we again find a clear difference based on language dorninance. While all of the children use uninflected German nouns without articles in their Turkish, the two active bilinguals TT 16 and TT 5 also use

-

Table 6 . Louii-ivord incorporarions in notol nnd verh plrrases h ) ~ foitr- 10 five-year olds (combined resulrs over one year of observarion.r in K ITA = prqjccr)' 9

Child/code/age German ekmenrs in Turkislr Turkish elemenrs in German 9 Noun phrases Verb phrases Noun phrases Verb phrases .

0 Art + Dt. noun+ Dt. Det. Dt. element Dt. verb 0 Art+ Dt. Det. Tk. element Tk. verb Dt. noun Tk. inflection +Dt. noun +Tk. verb Tk. noun +Tk. noun + Dt. verb

a. Sumrnarized frorn PfaRet al. (1988; 10 ab, Table 2).

Turkish in contact with German 1 I 1

German nouns with Turkish case inflections as in (1) and (2), a type of usage not found'in the speech of the German-dominant TD 2. Further, only TD 2 mixes whole German noun phrases (article+noun) into her Turkish.

(1) Int: kar yagdigi zaman ne yapiyoruz? (KITA TT 5m TK 4; 4) -'when it Snows what do you do?'

TT5: Schneemann bu [pointing at picture]. 'snowman this.'

(2) Vorschule'ye gitcem, ondan sonra okula (KITA TT 16f TK 4; 9). '1'11 go to preschool, then to school.'

Incorporation of German elements into verb phrases in Turkish also differs among the three children. The two active bilinguals typically use Gennan adjectival or nominal elements (including infinitive verbs) plus a neutral Turkish verb (yapmak 'make, do', etmek 'make, do', or olmak 'be') which carries the tense and Person inflection as in (3) and (4). a strategy not used by TD 2, who simply uses inflected German verbs.

(3) Geschenk etti (KITA TT 16 TK 4; 9). 'he/she gave a gift.'

(4) tanzen yapiyo (KITA TT 5 TK 5; 0). 'helshe is dancing.'

In the speech of the eight- to nine-year-old children from the EKMAUS study, we can again See clear effects of contact with Germans in the mixing behavior. Table 7 presents an overview of mixing found in the interviews in Turkish and in German of the three bilingual children here (See Pfaff 1988a and 1990a for a more detailed analysis of a larger EKMAUS subsample).

In this context of child-adult structured interactions, the overall fre- quency of mixing is very low and, as was the case with the KITA children, more frequently involves incorporation of German elements into Turkish than Turkish into German.

Comparing the Pattern of mixing German into Turkish for the three EKMAUS children, we note that not only the frequency but also the syntactic categories involved in mixing are related to amount of contact with Germans at school or at after-school day-care Centers. Mixing by the girl from group A (with little contact with German peers) is limited

' to the syntactic categories noun and interjection, as in (5) and (6):

( 5 ) bi Indianer degisik (EKMAUS A S2 I 1 f 9). 'one Indian is different.'

1 12 C. W. Pfafl

Tablc 7.. Loan-ward usage and rode swirrliing b~ eighr- ro nine-~eor-olds in EKMAUS sfu*

Gcnnan elerncnts in Turkish

Total Nurnkr . Approx. Nouns Verbs Adjcc- Inter- Adverbs Tclling words of pcnxni t iva jections/ time

mixing Gennan conjunc- Child instanca tions

' (6) ach so! burdan mi. basliyo? (EKMAUS A S2 I 1 f 9). 'oh! it Starts from here?

while the children from groups B and C mix in a wider variety of structures including verbs, prepositional phrases, calendrical expressions, as in (7)-(9):

(7) Peter schieben yapiyo (EKMAUS B S6 04 f 8). . 'Peter pushes.'

(8) hepsi im Kreis top da (EKMAUS B S6 04 f 8). 'they're all in a circle, the ball too.'

(9) Saat nach drei (EKMAUS C S8 01 m 8). 'after three o'clock.'

These differential Patterns, which hold also for the larger subsample reported in Pfaff (1 988a, 1990a) parallel the findings reported by Boescho- ten and Verhoeven (1985) for Turkish migrants in the Netherlands but with an important differente. While the children in their study mix a wide variety of Dutch elements into their Turkish, as do our group B (and some C) children, the Pattern shown by our group A children is character- istic of first-generation adults in the Netherlands. This finding reflects the extent to which sorne Turkish children in Berlin are isolated irom intcrac- tion with German Peers.

5. Case rnarking

Turning now from the lexical aspects of language maintenance and attri- tion to the effects of contact on morphosyntax, we first consider case marking in Turkish (see Pfaff and Savas 1988 for more details). As we noted, we expect case marking to be a robust System, unlikely to break \ 'i

down in the early Stages of language loss. Indeed, this seems to hold true

Turkish in conracr wirh German 1 13

Table 7 (continucd rrom top oC page I I?).

German elernenis in Turkish Turkish elements in Gennan

Ariicle Adjec- Mixed Prcposi- Verb Sentcnce Toial Nurnber Approx. + tive + cornpound tional phrase words ol percent noun noun phrase mixing Turkish

Child instances

in general for the bilingual children in Berlin, no matter what their age o r extent of contact with German.

Quantitative analysis of the expression of case on noun phrases over the first year of the longitudinal KITA project, given in Table 8, shows the frequencies of usage and the percentage nonstandard use for unin- flected (nominative and nondefinite accusative) and inflected nouns.

Note first that the frequencies of both uninflected and inflected nouns is considerably lower for T D 2, the strongly German-dominant child of a mixed marriage, than for the other two children, reflecting her avoidance of Turkish by resorting to German as noted in section 4 above. For TT 5, who is only slightly German-dominant, the frequency of inflected nouns is lower than for TT 16, the strongly Turkish-dominant girl.

The percentage of nonstandard case marking is low for all children, but considerably higher for inflected cases than for uninflected nomi- natives. This indicates that there is little or no tendency t o lose the inflectional System in favor of some more analytic means of expressing case relations, but rather that some nonstandard substitutions o ~ c u r . ~

In fact, analysis of the type ofsubstitution errors provides more Support

Table 8. NP rase marking in Turkish by Jour- tofiie-~ear-olds (KITA)

Uninflected N P s Inflecied NPs

N o m % n Acc Dat Loc Abl Inst Genj % Child n nstd Poss nstd

for.our finding that case marking is essentially intact in these young bilinguals. Errors are limited to peripheral semantic frames rather than representing prototypical case relations (See Zimmer 1988) and are paral- leled by errors made by Turkish monolingual children, as noted by Ekmekqi (1979, 1986). In (10) we find the use of accusative inflection rather than standard unmarked nominative for the subject of a passive verb:

(10) arabalari koyuluyo (KITA TT 5 m T K 5; 0). std: arabalar 'the cars are put in'

and in (11) we find dative rather than the standard locative for an expression of place in such a genenc Statement:

( I I) dügüne Eis var (KITA TT 5 m T K 5; 0). std: dügünde 'there is ice cream at the wedding'

I t is possible that in ( 1 I), which also contains a German noun. Eis, we may have an instance of syntactic transfer from German, which. lacking a locative case, would express this relation with a dative. This example would then Support the position espoused by Clyne (1988), who suggests that syntactic convergence often goes hand in hand with code switching and loan-word incorporation.

The results for the eight- to nine-year-olds from the EKMAUS study are parallel to those from the KITA study. Table 9 shows the realizations of case relations for agent; patient, recipient, location, etc.. elicited in the experimental task Actiotls, in which the children described events which the interviewer acted out with toys, for example, the boy pushes the car, the girl jumps over the clothespin (See below, section 7, and Pfaff et al.

Table 9. N P rase morking in Turkis/i by eighr- ro nitie-year-olds IEK.bl.4 L?S arfions sefs 1-5)

UninRected Inflected

Child n Norn. Acc. Dat. Dat.+ 'away' 'over' O/O

post 'from' Abl. + nstd pos. Abl. posr pos.

Turkish in corlracr wirh Gernzan I 15

1985: 48-53; PfafT and Savas 1988: 5; for further discussion of the methodology).

Like the KITA children, these EKMAUS children have all essentially maintained control of the case-marking system of Turkish. Even for the group B girl, whose overall percentage of nonstandard realizations reaches 15 percent, the types of nonstandard case marking are rather restricted. Case marking is correct in prototypical environments both with and without postpositions, as in (1 2)-(14):

(12) arabay~: the def. acc. object of schieben yapiyo (EKMAUS B S6 04 f 8). 'pushes the car'

(13) Wäscheklammer'i: the def. acc. object of aliyor (EKMAUS B S6 04 f 8). 'takes the clothespin'

(14) Wäscheklammer'in üstünden hopluyo (EKMAUS B S6 04 f 8). 'jumps over the clothespin'

Her difficulties occur when she attempts to refer to one of several similar or identical participants usin; more complex modified NPs. Thus in (1 5) she uses a formally correct but semantically anomalous possessive/ genitive construction, rather than the derivational -11 (See below, section 7.2).

(15) Indiana'nm pembesi (EKMAUS B S6 04 f 8). std: Indiana'nm pembelisi 'the pink dressed Indian'

In (16), where the nurse, one of three identical nurses in this Acrions Set, is the recipient, the dative marking is omitted or replaced by the genitive, a usage which would be appropriate to a compound noun:

(16) bir tane hemsiresi (EKMAUS B S6 04 f 8). std: hemsirelerden bir tanesine 'to one of the nurses'

Case-marking errors such as the above are typically limited to environ- ments which are structurally and semantically complex. Again, as for the KITA children, there is no general erosion of the case-marking system or replacement by analytic devices.

6. Number marking

The results for number marking can be summarized very briefly. The system, illustrated in section 3.3 above, appears to be essentially intact

I16 C. ,W. P m

for the KITA and EKMAUS children. Table 10 shows the realization of plural on nouns in plural quantified NPs for the eight- to nine-year-olds from the EKMAUS study.

Table 10. Plural marking on nouns in plural quanrified noun phrases -

Child Marked Unmarked Percent plural' plural nonstandard

A S 2 I I f 9 0 5 B S 6 0 4 f 8 2 20 9 C S801 m 8 0 29 0 D S 7 0 9 m 9 0 7 0

a. For.Turkish. marked plural is nonstandard. b. Percentages based on five or fewer instances are parenthesized.

As Table 10 shows, only the girl from group B with strong contact with German, who, as noted above, prefers German, has any instances of overmarking of plural in Turkish: 2 of 22 instances (9 percent). Note, however, that only one of these instances, given in (17), is in a simple NP context. The other instance occurs in a complex modified NP, (18), a structure which will be discussed further in section 7.3 below:

(17) qok cicekleri yedi (EKMAUS B S 6 04 f 8). 'he ate a lot of flowers'

(18) iki tune arka ayakkabilarinin (EKMAUS B S6 04 f 8). 'its two back shoes'

Such nonstandard plural usage in complex NPs accounts for the majority of nonstandard marked plural nouns for the larger sample of EKMAUS children, reported in Pfaff (1988a: 67-70). It is likely that the cause of such errors is not simply direct transfer from German but that complexity and processing demands also play a role.'

7. Modification

Three types of modification were presented in section 3.4 above: unin- flected adjectives preposed to the noun; a more complex derivational form with -!I; and a syntactic means of embedding sentential modifiers as participial clauses preceding the noun.

These structures, or the contexts in which they would be likely to occur, were systematically investigated in the EKMAUS study, where children needed to distinguish among similar but not identical items. simple adjectives, for instance, were elicited with toys which differed only in

Turkisli in contact with Gerntan 1 1 7

color (kirmrzr top 'the red ball' vs. mavi top 'the blue ball'). Partial diKerences involving color frequently elicited -11 (pembe elbiseli krzilderili 'the pink-dressed Indian' vs. yesil elbiseli krrrlderili 'the green-dressed Indian'). Participial clauses with subject heads formed with -En occurred in contexts where the diKerences involved an animal's stance (dört ayakta duran at 'the horse standing on four legs' vs. iki ayakta duran at 'the horse standing on two legs').

Table 11 provides an overview of modification for the eight- to nine- year-old EKMAUS bilingual children from groups A, B, and C and the monolingual control child from group D. The results are given qualita- tively rather than quantitatively because the number of instances of each context was low and because, particularly for the embedding structure, children often had several false Starts or required prompting: in Table 11, 'yes' indicates that the child spontaneously produced the s t r ~ t u r e in question; 'yes, but' indicates spontaneous use, but at least one instance of nonstandard structure or semantic anomaly; 'no, but' indicates that the child did not produce the structure spontaneously but did so after modeling by the interviewer; and 'no' means the structure was not used at all.

7.1. Adjective + noun

As shown in Table 1 I, all the children in the subsample maintain the simple adjective+ noun structure as predicted. Although we note some uncertainty about the Standard color-term reference (for example, 'blue' for 'red'), this is not our concern here.

Table I I . Noun-phrase nto(li/icarion by eighr- ro nine-yeor-olds in EKMAIiS projecr'

Child code Adj -11 -En participle

A S 2 I I f 9 Yes Yes no B S6 04 f 8 Yes no. but no, but C S 8 0 1 m 8 Yes Yes Yes D s 7 09 m 9 yes yes yes

a. 'yes' =child's spontaneous production of structure. 'yes. but'= spontaneous production with at least one instance of nonstandardness. 'no' =no use of structure at all. 'no, but' = structure not produced spontaneously.

7.2. Derivarional -11

The use of the derivational morpheme, -[I, is Standard in the utterances of the EKMAUS A and C children shown in (19) and (20), as well as in the monolingual child's speech.

(19) pembe elbiseli (EKMAUS A S2 1 1 f 9). 'the pink-dressed (one)'

(20) pembe elbiseli kizilderili (EKMAUS C S8 01 m 8). 'the pink-dressed Indian'

However, this structure is not produced by our group B girl, who, as we saw in (15) above, repeated here as (21), employs a possessive/genetive structure which is clear in its intent but ungrammatical in this context and semantically anomalous if taken literally:

(2 1) Indiana'nm pembesi (EKMAUS B S6 04 f 8). 'the pink Indian' lit. 'the Indian's pink'

While this group B child does not seem to have the derivational -/I structure (or did not produce it in a likely context), some of the younger children from groups A and B in a larger EKMAUS subsample (reported in Pfafi 1989) produce the derivational suffix in modifying phrases but have difficulties with its placement, incorrectly attaching it to the head noun, as in (22) and (23):

(22) kirmizi Indianer'li (EKMAUS A S2 06 f 7). 'red Indian'

(23) yeSilIndianer'li (EKMAUS B S5 01 f 6). 'green Indian'

Such ungrammatical placement of the derivational morpheme is not found in the speech of the group C children, who are all older and have attended at least one year of school in Turkey, or in the speech of the kindergarten or primary school children from the monolingual control group. Thus it appears that the derivational structure is one which is affected in the process of language attrition here.

7.3. -En parriciples

Participial structures are rare among the older EKMAUS bilinguals. The monolingual control group D boy and the group C bilingual who had

Turkisl~ in contact rvith German 119

attended school in Turkey before immigrating to Berlin produce Standard subject relative structures, as in (24) and in the first clause of (25):

(24) kurt yaian inege topu veriyor (EKMAUS D S7 09 rn 9). 'the wolf gives the ball to the cow which is lying down.'

(25) oturan inek, oturan deiil, o ayaktaki yani (EKMAUS C S8 01 m 8). 'the sitting cow, not the sitting (one), that is the one which is standing.'

but also use other paraphrases as well, as in (26) and in the second clause of (25) above:

(26) kurt ayaktaki inegin üstüne hopladi (EKMAUS C S8 01 m 8). 'the wolf jumps onto the cow which is standing.'

For the group A and group B children, however, there is no evidence that the participial structure has been acquired. The group A girl a t first avoids participial modification, substituting reference to size, a secondary distinguishing characteristic of the two cows, then produces a reference to its stance in (27):

". (27) kücügüne, ayaktakine (EKMAUS A S2 11 f 9). 'to the small one, to the standing one'

She uses a further paraphrase with a finite verb to refer to the rearing horse in (28):

(28) aj~aktaki assadan kollarini kaldirdi (EKMAUS A S2 1 1 f 9). 'the standing one, 0 raised its arms from the bottom'

The group B girl also uses finite verbs but with the subject postposed, perhaps her closest approximation to the preposed participial structure in (29)-(30):

(29) iki tune ayakta duruyo o at (EKMAUS B S6 04 f 8). 'that horse is standing on its two feet'

(30) o duruyo ya, o ineiJe (EKMAUS B S6 04 f 8). 'that one, it's standing, you know, to that cow'

using the discourse particle ya paraphrase noted by Slobin (1986: 279) in the Speech of much younger children. Finally, after prompting by the interviewer, she attempts a participial structure in (31) but breaks it off to begin a new sentence with a finite verb.

(3 1) o ayakta duranrn inegin - inekten atliyo (EKMAUS B S6 04 f 8). 'that cow which is standing - 0 jumps from the cow'

Since alternative paraphrases are frequent in everyday Speech and these children receive no instruction in Turkish language skills in school, it is very unlikely that the participial structure has been 'lost' by these A and B bilingual children, but rather that it has never been acquired.

8. Pronominal vs. null anaphora

The final aspect of nominal reference we examined is pro-drop phenom- ena, specifically whether contact with German. a non-pro-drop language, leads to overuse of pronominal reference rather than null anaphora, which is Standard in clear discourse contexts in monolingual Turkish. As was mentioned earlier, the use of anaphoric pronouns is optional, but, as pointed out by Enc (1986), Erguvanli-Taylan (1984), and Slobin and Talay (1985), among others, use vs. nonuse of pronouns in ongoing discourse is not equally appropriate. When the reference does not switch or when the intended referent is clear from the context, it is more appropriate to omit the pronoun.

To address this issue, we examine first- and second-person reference in the context of pretend dialogs with dolls in the Plajrooms game and third-person reference in the descriptions of events acted out with dolls in the Aciions game.

First- and second-person contexts are illustrated in (32)-(35), which give the realizations of the Same request for the four EKMAUS children: the child wants to borrow the yellow ball from the boy doll:

(32) Orhan, Sari topu alabilir miyim? (EKMAUS A S r 1 1 f 9). 'Orhan, may I have the yellow ball?'

(33) Peter, sen bana topu verir misin? (EKMAUS B S6 02 f 8). 'Peter, will you give me the ball?'

(34) Ahmet, senin topunu alabilir miyim? (EKMAUS C S8 01 m 8). 'Ahmet. can I have your ball?'

(35) Ahmet, topu ver! (EKMAUS D S7 09 m 9). 'Ahmet, give me the ball!'

The utterance of the group B girl in (33) contains the discourse-redundant second-person subject pronoun sen and first-person indirect-object pro- noun bana. As shown in Table 12, she typically uses such redundant first- and second-person pronouns, while these independent pronominal forms are never used by the children of the other groups.' The group A and C children rely on verb agreement, while the monolingual group D child omits all overt reference to Person by using the imperative.

The results for third-person reference point in ,the Same direction as ,

Turkish in contact witli German 12 1

Table 12. ~ i r s i - atid secund-person relerence in plavroom' producrion: pronuun und Zero

anaphora (pertrnr)

Child/code n YO YO pronoun 0 anaphora

a. Repeats interviewer four of the five instances.

those for first and second Person. Potential contexts for third-person reference occur primarily within linked predications in the individual Actions to be described. Examples (36)-(39) illustrate the realization by each of the EKMAUS children for an event which consists of a sequence of two parallel actions on two identical objects of the same type by different agents (Set 3, event 6) (subscripts ,? j, ,, etc., indicate coreference or disjoint reference): 'the girl pushes the car then the boy pushes another car.'

(36) Arzu, gitti 0, arabayij itti. Orhan, geldi o, da 0, ters taraftan itti (A S2 1 1 f 9). 'Arzu, went Oi pushed the carj. Orhan, came, he, too pushed 0, from the other side'

(37) Steffii arabayij schieben yapiyo, Peterk arabayi, schieben yapiyo (EKMAUS B S6 04 f 8). 'Steffi, pushes the carj, Peter, pushes the car,.'

(38) Aysegül ve Ahmet, arabalarij ittiler (EKMAUS C S8 01 m 8). 'Aysegül, and Ahmet pushed the carsj.'

(39) Ayse, arabayij itekledi, Ahmet, de arabayi, itekledi (EKMAUS D S7 09 m 9). 'Ayse, pushed the carj, Ahmet, too pushed the car,.'

Overt expression of the object noun in the second predication represents Standard Turkish usage, in which the overt element corresponds to the disjunct reference of the two cars. Zero anaphora is also pragmatically appropriate in this context because the referents were present in the extralinguistic environment.

A more complex example involving contexts for Zero anaphora of both I

subject and object is shown in (40)-(43), which give the realization of a r;

complex action involving several predications of the same agent and object (set 5, event 19): 'the Indian with the pink dress takes the clothespin

from the Indian with the red dress, jumps over it, and brings it to one of the nurses':

(40) pembe elbiseli, aldi k~rmizinin~ yanindan mandali, sonra Oi O, üs- tünden hopladi Oi 0, 0, getirdi (A S2 11 f 9). 'the pink dressed (one), took the clothespin, from beside the red (dressed ~ n e ) ~ then 0, jumped over O, 0, brought 0, (to O,).'

(41) simdi Indiana'nin pembesi, Wä~cheklammer'in~ üstünden hopluyo 0, gine arkaya geliyo, Oi Wä~cheklammer'i~ aliyor Oi bir tane hemsir- esi, Wä~cheklarnmer'i~ venyog (B S6, 04 f 8). 'now the pink Indian, jumps over the clothespinj (4, Comes back again, Oi takes the clothespinj 0, gives the clothespinj to one of the nurses,'

(42) pembe elbiseli kizilderili, mandalij aliyor Oi mandalij yere birakti Oi Oj aldi gene Oi Oj hemsire, götürdü Oi Oj gene aldi (C S8 01 m 8). 'the pink dressed Indian, takes the clothespinj Oi left the clothespinj on the ground 0, took Oj again 0, brought Oj to the nurse, 0, took Oj again.'

(43) o yesil kiz, mandalinj üstünden hopluyor, 0, zipliyor, 0, mandalij alip 0, Oj hemsirenin, yanina götürüyor (D S7 09 m 9). 'that green girl, jumps, 0, hops over the clothespinj, Oi takes the clothespinj Oi brings Oj beside the nurse,.'

The results for third-person singular subjects and direct objects are shown in Table 13.

The figures in Table 13 show that the group B girl is the only one of the four children to produce any pronominal reference for subjects and that her rate of pronominal reference to direct objects is the highest. At the Same time, she has the lowest rate of pro-dropping.

The effect here is rather slight in comparison to the results for first- and second-person reference discussed earlier. This can be attributed to the elicitation context we analyze here (the Actions game), which requires the child not to produce connected discourse, but rather to describe each

Ta ble 13. Third-person rejerencejor subjects und direci objects in octions (seis 1-5): noun phroses, pronouns, und 0 anaphora (percent)

Child/code Subject Direct object n % N P %Pro % 0 n % N P %Pro '100

Turkish in contact with German 123

event immediately after it is acted out by the interviewer, thus pragmati- cally licensing a 'new start' for each event within a Set of actions with the Same toys. Other contexts, particularly extended narratives (Rehbein, personal communications; Pfaff and Cakarcan 1989) show higher fre- quencies of third-person pronominal reference in bilingual's than in mon- olinguals' speech.

Increased use of pronominal anaphora in conjunction with a decline in the rate of pro-dropping can be interpreted as transfer from the non- pro-drop contact language, German. This may indeed be a contnbuting factor but need not be the only source of this phenomenon. As Givon (1976: 154) has suggested, the redundant overuse of discourse machinery may be an effective communicative strategy when the communicative system is relatively frail, as he notes is the case with early child language. Communication in the first language by bilingual children who are domi- nant in the second language appears here as another setting for communi- cation using a relatively frail system which can lead to such unnecessary repetition.

9. Conclusions

The results of our study document certain aspects of the ongoing pro- cesses of language development in individuals and of the language vari- eties used by migrants from Turkey in Berlin (West). Here, as in other cases of language shift and loss, we have Seen that the major determinants of the degree of developmentlattrition of the ethnic-minority language are the social factors which provide the framework ofcontact and interac- tion with native speakers of Turkish and German. Particulars of the linguistic varieties of Turkish which are emerging here, howeuer, are constrained by the basic typological characteristics of Turkish and the psycholinguistic simplicity o r complexity of the grammatical subsystems under investigation. Thus it is the transparent, regular inflectional mor- phology which proves to be robust and resistant to attntion, while the more complex and opaque areas of derivational morphology and syntax are subject to attrition, characterized by avoidance, morphological substi- tution, placement errors, and syntactic paraphrases.

Thus, we find that not only is the extent of mixing of German elements into Turkish quantitatively greater for children with more contact with Germans but the syntax of mixing differs as well. All the bilingual children mix German nouns (with Turkish inflections where appropriate) and interjections into their Turkish, but only those with more contact with

124 ' C. W. Pfif f

German mix in verb phrases and adpositional phrases a s well and show instances of intersentential code swltching.

As for the specific morphosyntactic features of nominal reference we examined in detail, the following generalizations can be made:

The inflectional morphology for case and number are essentially intact. The small amount of nonstandard usage found appears to be restricted to nonprotypical environments in complex and peripherial structures. a Pattern which is similar to that noted for monolinguals, though it may continue to a later age here.

Derivational morphology, to the extent that it is revealed here, seems to be somewhat weaker. Modifiers with -11 were avoided by the German- dominant child we examined in detail and sometimes misplaced by other younger bilingual children.

The complex syntax required for embedded sentential modification is clearly late in appearing in the second-generation migrant children. and, for some, it may be entirely lacking. The function of sentential modifica- tion is carried out by other paraphrases, which are frequent in Standard spoken Turkish. Whether the participial forms will eventually be acquired by these children o r whether the syntax of Turkish will undergo chanee. as it has in previous periods of contact, remains to be seen.

The pragmatic and stylistic features of the contact varieties of Turkish which may reflect linguistic convergence show up not as grammatical 'errors' but as.overuse of one of the possible syntactic alternatives in a given discourse context. Thus, in our analysis of nominal reference in requests and in descriptions of sequences of events acted out with toys. we find some evidence of more frequent use of explicit pronominal rather than null reference for the most German-dominant child investigated here. This phenomenon is Open to two interpretations, which, it should be noted, are not mutually exclusive: it may be seen as, on the one hand. evidence for transfer of the 'non-pro-drop' Parameter settine from Ger- man or, on the other hand, as a communication strategy which provides greater redundancy when the communicative System, the ethnic-minority language, has become frail through attrition. Resolution of this question will require further investigation and comparison of more discourse contexts for Turkish monolinguals with bilingual children with various Patterns of language dominance.

As we have seen, the situation of Berlin (West) provides a very complex sociolinguistic setting for maintenance and attrition of Turkish and other ethnic-minority languages. In particular, the children of migrants from Turkey may be confronted with one o r more of a wide variety of quite different conditions for the acquisition of their two languages. As a result of the educational policy which allows Ausländerklus.sen, those from

Turkish in contact with Gerrnan 125

neighborhoods with a high concentration of Turkish children and adoles- Cents rnay experience a linguistic environrnent in which Turkish rernains the dominant language into the second and third generation, while others are rnuch further advanced on the continuum toward language shift. The fact that rnother-tongue instruction in Turkish is not generally available a t the initial pnmary level and Turkish in place of a foreign language is chosen by only a rninority of pupils at the late prirnary and secondary levels rneans that the input in Turkish for all children is largely restricted to informal discourse from adults and other children rather than including rnore formal registers.

Sorne of the effects of these conditions on the lexicon, morphology, and Syntax of Turkish spoken by rnigrant children here have been shown in this paper. Much rernains for further investigation both on the specifics of Turkish in contact with Gerrnan and on the rnore general problern of linguistic change in contact settings. Cross-linguistic and cross-cultural cornparisons, as facilitated by international conferences such as the one a t which this paper was originally presented, are essential to the achieve- rnent of the larger goal of understanding the interplay of linguistic and social factors in this highly cornplex field.

Freie Universitär Berlin

This paper was presented at the Workshop and Conference on Maintenance and Loss of Ethnic Minority Languaxes. Noordwijkerhout, The Netherlands Aug. 28-Sept. 2, 1988. The work represented in this paper was done by a team of researchers over the years 1983-1988. Those directly involved with the data collection and analysis of the Turkish varieties presented here are (in alphabetical order) Murat Ceyisakar, Ogün Cakarcan, Filiz Kardarn, Dina Kohen. Tülay Savas, Chnstine Williams.

Special thanks are due to Idil Temuqin, who was the director of the preschool day- care center in which some o f t h e EKMAUS interviews and all of the KITA study were conducted.

We are also indebted to Özden Ekmek~i and Zeynep Benli Tor arranging and conduct- ing the control-group interviews Tor the EKMAUS study in Ankara.

Funding was provided by the Freie Universität Berlin 1983-1986 and the Deutsche Forschungsgemeinschaft since 1987.

The terms Berlin (West) and West Berlin which applied at the time discussed are retained in this paper. The recent political events can be expected to have demographic consequences which will aKect the sociolinguistic settings significantly. For instance, since the children we studied lived in neighborhoods which were peripheral in West Berlin but which are now in the center of the reunified city. there will be increased opportunities for contact with Germans from the former GDR.

I . This preschool setting is nor typical. It is based on the premise that giving equal rights

126 C. W. Pfaff

to both languages will not only prove beneficial for the linguistic development of Turkish children. but that it will also support - or even help to create - better attitudes toward minority languages and their speakers (See Kardam et al. 1988).

2. As Sally Boyd noted in her comments during the workshop, a full investigation of this hypothesis would require comparison of the rates and nature of loan-word integration in other contact Situations. Given the wide variety of migrants in G e p a n y and other northwestern European countries, such a study is feasible but clearly beyond the scope of the present paper.

3. Rik Boeschoten (personal communication) points out that in certain marginal construc- tions for groups acting together, the plural marker, -/er is permitted with plural quantifi- ers. for instance in titles such as üq silahjörler 'The Three Musketeers' (see Johanson 1977 Tor discussion). These are not the constructions under discussion in the present paper.

4. In this paper we follow the standard orthographic conventions Tor Turkish. Capital letters are used as Cover terrns Tor vowel and consonant harmony variants as follows: E =e.a: I =i,i.u.ii; D=d,t .

5. Our studies indicate that. despite the occurrence of sentences in which lexical iterns from both languages occur, these bilingual children have clearly developed two separate Systems. It is thus appropriate to speak of 'discourse in Turkish' and 'discourse in Gerrnan' from the psycholinguistic standpoint of the child as well as from the interac- tional contexts. Analysis of the Speech of the younger KITA children (Pfaff et al. 1988, Pfaff 1990~) indicates that such Separation of Systems is present from the outset. supporting the position taken in other recent studies of bilingual language acquisition. such as Meisel (i.p.).

6. A possible exception here is the genitiveipossessive construction. marked both on the possessor NP (with genitive inflection or a possessive pronoun) and on the NP which is possessed (with possessive inflection). as in Ali'nin para-si 'Ali's money' beri-irpi para- 111 'rny money'. We find a tendency to omit the second inflection. the possessi~e marking on the possessed noun, in exarnples such as ben-ini para-0 iSar ' I have money' (KITA TT 5m 4; 10). This type of error is also made by a German girl, DD7. in her Turkish second language: hen-im anne-0 nerde? 'where is my mother?.

As Ayhan Aksu-KOG (personal communication) notes, the ornission of such possessive inflections represents more reliance on analytical marking of grammatical relations than in standard Turkish. However. we observe that this process appears not tobe generalized to other cases but is restricted to the genitiveipossessive with personal pronouns where the inflection is redundant. It rernains to be seen whether 0 inflection is characteristic of possessives without overt pronouns and. crucially, whether some analytic marking develops in such environrnents and in other cases where the infleciiori is the Sole rnarker of the grammatical relation.

7. It is of interest to note, however. that the Turkish rule requiring nouns to take the singular form in NPs with plural quantifiers may be transferred to German by children who are dominant in Turkish. For discussion see PFaff (1988a: 70. 1988b).

8. Occasionally, however, the equally redundant possessive pronouns are used, as in (34) .renin ropurru. which is marked by the genitive on the possessed noun rop-un-u 'your ball' - der. acc., as well as by the possessive pronoun senin 'your'. Whether this is related to the tendency toward loss of the genitive rnentioned in note 5 is a topic Tor further investigation.

9. The nonstandard structures in (41). lndianahin pembesi and bir iunc hemgiresi. are discussed in sections 5 and 7.2 abovc.

Tlrrkish in conracr tclirh Gernjan 127

References

Aksu-Kq, Ayhan, and Erguvanli-Taylan, Eser (eds.) (1986). Proceedings of 11e Turkish Linguisrics Conference, August, 9-10. 1984. Istanbul: Bogazici Universitesi Matbaasi.

-, and Slobin, Dan I. (1985). Acquisition of Turkish. In The Cross-Linguisric Slu& of Longuoge Acquisirion, vol. I. Dan I. Slobin (ed.). 839-878. Hillsdale, NJ: Erlbaum.

Blaschke. Jochen, and Ammann, Birßt (1988). Kurden in der Bundesrepublik Deutschland: Ihre soziale und kulturelle Situation. In Kurden: Allrag und Widersrond, Yayla Mönch- Bucak (ed.), 90-99. Bremen: Eigenverlag der Herausgeberin.

Boeschoten. Henrik, and Verhoevan, Ludo (1985). Integration niederländischer lexikalischer Elemente ins Türkische: Sprachmischung bei Immigranten der ersten und zweiten Gene- ration. Linguisrische Berichte 98, 347-364.

-, and Verhoeven, Ludo (eds.) (1987). Sfudies on Modern Turkislt. Proceedings of rhe Third Conference on Turkirh Linguisrics. Tilburg: Tilburg University Press.

Cakarcan. Ogün (1988). Erstsprache eines in der Zweitsprache dominanten Kindes. Unpub- lished Hauptseminararbeit, Freie Universität Berlin.

Cakir, Mehmet Yasar, and Meyer. Christian (1988). Ansätze eines konfrontativen türkisch- deutschen Sprachförderunterrichts. Deursch lernen - Zeirschri/r/ur den Sprochunrerrichr mir ouslöndkchen Arbeitnehmern 13 (2), 46-59.

Clyne, Michael (1988). Linguistic and sociolinguistic aspects of language contact. mainte- nance and loss: towards a multifacet theory. Paper presented at the Conference on Maintenance and Loss of Ethnic Minonty Languages, Noordwijkerhout Aue. 31-Sept. L.

Cummins. James (1984). Bilinguolism und Special Educorion: lssues in Assessmenr und Pedagogy. Clevedon: Multilingual Matters.

Ekmekci. Özden F. (1979). Acquisition of Turkish: a longitudinal study on the early language development of a Turkish child. Unpublished doctoral dissertation. University of Texas, Austin.

-(1986). The developmental errors in the pre-school Turkish children's speech. In Proceed- ings of rhe Turkish Linguisric Conference Augusr 9-10 1984, 249-267. Istanbul: Bogazici Universitesi Matbaasi.

Eng. Müwet (1986). Topic switching and pronominal subjects in Turkish. In Srudies in Turkish Lingui.~rics, Dan Slobin and Karl Zimmer (eds.), 195-208. Amsterdam/Philadel- phia: Benjamins.

Erguvanli. Eser (1984). Tlre Funcrion of Word Order in Turkish Gramnior. Berkeley: Univer- sity of California Press.

Givon, Talmy (1976). Topic. pronoun and grammatical agreement. In Subjecr und Topic, Charles Li (ed.), New York: Academic Press.

Grönbeck. K. 11936). Der riirkisrlre Sprochbou. Copenhagen: Levin and Munksgaard. Hankamer. Jorge. and Knecht. Laura (1976). The role of subjecthon-subject distinction in

determining the choice of relative clause participle in Turkish. In Horvord S~udies in Synrox und Semanrics 2, Jorgen Hankamer and Judy Aissen (eds.). 97-221. Cambridge, MA: Haward University Press.

Hyams, Nina M. (1986). Longuage Acquisirion und rhe Theory of Paromerers. Dordrecht: Reidel.

Johanson, Lars (1977). Bestimmtheit und Mitteilungsperspektive im türkischen Satz. Zeirschri/i der Deurschen Morgenlöndrjchen Gesellschofr, Supplement 312: 1 186-1203.

Kardam, Filiz, Cinar, Safter, Temucin. Idil, and Pfa& Carol W. (1988). Migrants from Turkey in West Berlin: Status, legislation, language rights and approaches to bilingual

ed.ucation. Paper presented at the Conference on Minority Rights and Minority Educa- tion, Cornell University. May 7-9.

Keskin. Anne (1988). Alphabetisierung in der Muttersprache. Deursch Lernen-Zeiuclirijl fur den Sprachunrerrichr mir ausländischen Arbeitnehmern 13 (2). 18-45.

Kornfilt. Jaklin (1987). Turkish and the Turkic languages. In The World's Major Langu- agues. Bernard Comrie (ed.), 619-644. London and Sydney: Croom Helm. -. (forthcoming). Current issues in Turkish Syntax. In Srare of ihe Ar! in Turkisli Linguis- rics, Henrik Boeschoten and Ludo Verhoeven (ed.).

Meisel, Jürgen (i.p.). Early differentiation of languages in bilingual children. In Bilingualism Across rhe Lijppan: I n Healrh and in Parhology, Kenneth Hyltenstam and Loraine Obler (eds.).

Nehr. Monika. Birnkott-Rixius Karin. Kubat, Leyla and Masuch, Sigrid (1988). In zwei Sprachen Lesen lernen - geh! denn das? Weinheim and Basel: Beltz Verlag.

Nilsson. Birgit (1985). Case marking semantics in Turkish. Unpublished dissertation. Stock- holm University.

Pfaff. Carol W. (1985). The problem ofplurifunctionality in bilingual language acquisition. Papers on Research in Cllild Language Deselopmenr 24, 95-103.

-(1988a). Linguistic and social determinants of TurkishIGerman bilingualism of migrant children in Berlin. In Diaspora Languages in Wesrern Europe. Lund: Slaviska Institutio- nen vid Lunds Universitet.

-(1988b). The development of nominal reference in Turkish-Grrman bilingual children. Paper presenied at Ist Hamburg Symposium on Multilingualism: Dimension of Bilingual Development. September 8- 1 1.

-(1989). Use of Turkish by migrant children in Berlin (West). Paper presented at the Second Workshop on Turkish in North Western Europe, Tilburg. January 23-21.

-(1990a). Mixing and Linguistic Convergence in Migrant Speech Communities: Lineuistic Constraints. Social Conditions and Models of Acquisition. Discussion paper Tor the second workshop of the European Science Foundation Network on Code-switching and Languagr Contact: Constraints, conditions and models, London.

-(1990b). Toward an lntegrated Approach io the Development o f ~Morphosyntactic and Communicative Compentence of Migrant Children: The Turkish of Bilingual Children in Grrmany. Paper presented at the Workshop on Ethnic Minority Languages in Europe. Tilburg. December 14.

-, Borde. Theda, Kohen. Dina, and Mügge. Ute (1985). Repvrr EKMAL'S-Teilprojekr: Linguisrisclie und kognirive Enrtc-icklung: Die Beziehung z~iischen Ersr- und Zti~eirspraclier- werb. Berlin: Freie Universität Berlin.

-, and Cakarcan, Ogün (1989). Nominalreferenz in Erzählungen im türkischen und deutschen Sprachgebrauch von mono- sowie bilingualen Kindern. Paper presented at the I. Türkisch-Deutschen Sprachwissenschaftlichen Symposium. March 20-22. Ankara.

-, Kardam, Filiz. VoD, Jutta. Freesemann. Heike, S a v a ~ . Tülay. and Thiel, Martina (1987). Arbeitschericht I on DFG-Project Narürlicher bilingualer Spraclierwerb von KIT4-Kin- dern: Krippenuller bis ersre Grunhchuljahre.

-. Kardam. Filiz. VOD, Jutta, Freesemann, Heike. S a v a ~ , Tulay. and Thiel, Martina (1988). Arbeitsbericht I1 on DFG-Project Nurürlicher bilinguuler Sprucherii,erb von KITA-Kin- dern: Krippenulrer hic ersre Grundcchuljahre.

-, and S a v a ~ , Tülay (1988). Language development in a bilingual setting: the acquisition of Turkish in Germany. Paper presented at the Fourth.Conference on Turkish Linguis- tics, August 17- 19. Ankara.

Phinney, Marianne (1987). The pro-drop Parameter in second language acquisition. In

T~trkish in cotttacf iiitlt German 129

Purun~eicr St,rrit~g. Thomas Roeper and Edwin Williams (eds.). 221-238. Dordrecht: Reidel.

Pritsak. 0 . (1963). Das Alttürkische. In Handbuch der Orientalisrik. 1.5:l. B. Spuler (ed.), 27-52. Leiden: Brill.

Senator für Schulwesen. Berufsbildung und Sport (1987/1988). Das Schuljahr 1987188 in Zahlen. Berlin.

S kutna bb-Kangas. Tove ( 1 984). Bilingualism und Special Educarion: Issues in Assessment und Pedagogy. Clevedon: Multilingual Malters.

Slobin, Dan 1. (1982). Universal and particulars in the acquisition of language. In Longuage Acquisirion: Tlie Srare o j rhe Ar!. Eric Wanner and Lila Gleitman (eds.), 128-170. - Cambridge: Cambridge University Press.

-(1986). The acquisition and use of relative clauses in Turkic and Indo-European lan- guages. In Siurlie.~ itl Turkisli Litgiiisrics. Dan I. Slobin and Kar1 Zimmer (eds.), 273-294. Amsterdam Philadephia: Benjamins. -. and Bever, Thomas G. (1982). Children use canonical sentence Schemas: a crosslinguistic study of word order and inflections. Cognirioti 12, 229-265.

-. and Talay. Ay~egül (1985). Development of pragmatic use of subject pronouns in Turkish child language. In Proceeditigs o j rlte Turkish Linguisrics Conjerence. Ayhan Aksu-KOG and Eser Erguvanli-Taylan (eds.). 207-228. Istanbul: Bogazici Universitesi Matbaasi.

-, and Zimmer. Karl (eds.) (1986). Siudies in. Turkisli Linguisrics. AmsterdamiPhiladelphia: John Benjamins.

Statistisches Landesamt Berlin (1988). Berliner Siarisrik. Berlin. Steinmüller. Ulrich (1987). Spracheniwicklung und Sprachunterricht Türkischer Schüler

(Türkisch und Deutsch) im Modellversuch 'Integration ausländischer Schüler in Ge- samtschulen.. In Hel_eö Thomas (ed.) Modellversuch Inregrarion ausländischer Schüler un Gesan~rschirlt~ti (1982- 1986). 207-315. Berlin: Pädagogisches Zentrum Berlin.

Thomas. Helga (1987). Zusömmenfassende Darstellung der wissenschaftlichen Begleitung zum Modell\.ersuch 'Integration ausländischer Schüler in Gesamtschulen' an der 2. Oberschule in Brrlin Kreuzberg und der I . Oberschule in Berlin-Tiergarten. In Helga Thomas (hrsg.) ;\foi~ellirrsuch Inregrurioti ausländischer Schüler iti Gesan~rscliulen (I 982- 1986). Berlin: Pädagogisches Zentrum Berlin:

&ar. Ali (1986). Die Stellung der ausländischen Lehrer in der Berliner Schule und die Kooperationshindernisse. In So:ialisorionsbedingungen ausländischer Kinder. 470-548. Materialien zur Lehrerfortbildung im Ausländerbereich (LiA) Er. 27. Berlin: Senator für Schulwesen. Berufsausbildung und Sport.

Underhill. Robert ( 1972). Turkish participles. Linguisric Inquiry 3, 87-99. -(1986). Turkish. In Srudies in Turkish Linguisrics. Dan 1. Slobin and Kar1 Zimmer (eds.).

Amsterdam: Benjamins. Verhoeven. Ludo. and Boeschoten. Henrik (1986). First language acquisition in a second

language submersion environment. Applied Ps~cholinguisrics 7, 241-256. -. and Boeschoten. Henrik (eds.) (1987). Siudies on modern Turkish. Proceedings o j rlie

rhird cot!ference on Turkish linguisrics. Tilburg: Tilburg University Press. White, Lydia (1985). The 'pro-drop' Parameter in adult second language acquisition. .

Lunguuge Leurning 35 (1 ), 47-62. Zimmer. Kar1 (1988). Semantic motivation in case assignment: another look a t Turkish

causatives. Paper presented at the Fourth Conference on Turkish Linguistics. Ankara, August.