geert driessen & michiel van der grinten (1994) see home language proficiency in the...

Pergamon Studieain E#u~z~onal Evalua~n, VoL 20, pp. 365-386, 1994

@ 1994 Elsevier Sci~co Ltd Pnnted in Groat Britain. All r i~t8 reserved.

0191-491X1941~26.00

0191-491X(94)00030-1

HOME LANGUAGE PROFICIENCY IN THE NETHERLANDS: THE EVALUATION OF TURKISH AND MOROCCAN BILINGUAL

PROGRAMMES-A CRITICAL REVIEW

Geert W. J. M. Drlessen and Mlchlel A. van tier Grlnten Institute for Applied Social Sciences, University of Nijmegen, Nijmegen, The Netherlands

Background

The Non-indigenous Population in the Netherlands

In the Netherlands it is possible to distinguish three categories of non- indigenous people. First, there are the migrants from former Dutch colonies, i.e., inhabitants of the former Dutch East Indies, the Moluccan Islands, Surinam, the Antilles, and Aruba. As a result of their ties with the (former) motherland of the Netherlands, these migrants were already somewhat acquainted with the Dutch language and culture. The second category are the so-called migrant workers. These mainly came from Mediterranean countries, such as Italy, Spain and later also from Turkey and Morocco. One characteristic all of these migrants have in common is their low level of education; in addition, their language and culture is very different from the Dutch one. This is particularly true for the Turks and Moroccans. The third category of non- indigenous people is made up of refugees from Eastern Europe, South-America, Asia and--more recently--Africa, the Middle East, and again Eastern Europe. This is a very diverse category both in terms of language and culture.

These three categories are made up of roughly 650,000, 500,000 and 40,000 people.l Within the second category, Turks and Moroccans make up the largest group with respectively 206,000 and 168,000 people. The Netherlands has a population of about fifteen million inhabitants, about 6% of the population is non-indigenous. Of the 119,000 non-indigenous children attending primary education, 38,000 are Turkish and 39,000 Moroccan. The total percentage of non-indigenous children in primary education is twelve per cent (data for the year 1990).

Educational Models

The arrival of non-indigenous children in Dutch education met with various educational approaches (Verhoeven, 1987). It is possible to distinguish two main

3 6 5

366 G. W. ,I, M. Driessen and M. A. van der Grinten

models in these approaches, namely monolingual and bilingual models. From the very start, a variant of the monolingual model, i.e., the submersion model, has frequently been used in primary education. In this model non-indigenous children are placed directly into Dutch classes, where they receive monolingual (Dutch) education. At some schools, this model has been adapted somewhat by taking the children from their classes at set times and giving them additional tuition in Dutch as a second language. In both cases, the learning of Dutch plays a central part and no attention is paid to the home language of the children. Children who have been in the Netherlands for a relatively short period of time can receive intensive instruction in Dutch for a maximum period of one year, with the aim of making it easier to talk to them in Dutch. In the submersion approach, occasional use is made of the home language; in these situations foreign teachers, especially Turks and Moroccans, are called in to assist the group teacher. In this situation, the function of the home language is a purely supportive one.

In bilingual models, the home language is used as well as Dutch. The extent to which the home language is used does vary, however. In a first variant of the model, called the transitional model, pupils are instructed in their home language during the first year. In the second year they are also given instruction in Dutch. At the end of that year, instruction in their home language is stopped and instruction continues in Dutch only. In this model, the home language is seen as a temporary aid to facilitate the learning of Dutch. A second variant is the simultaneous model. In this model, the children receive half of the instruction in Dutch and half in their home language, for a period of two to three years. After this period, a complete transition is made to Dutch. Although the home language is used as the language of instruction for a great deal of the time, the main emphasis is still on learning Dutch. This does not alter the fact, however, that it is also possible to use the simultaneous model to achieve a certain degree of functional biliteracy (language maintenance). In the Netherlands, transitional and simultaneous models have been applied at only a few schools. A third, very frequent variant of bilingual education is Home Language Instruction--HLI. 2 This variant has been around, unofficially, since 1967; since 1970 it has official status and has been co-financed and organized by the Dutch government. In principle, pupils can receive HLI throughout their entire primary education. Furthermore, there are also possibilities for receiving HLI at secondary schools. Since its introduction, there have been some changes in relation to the aims to be achieved (Lucassen & K6bben, 1992). Initially, HLI was mainly aimed at maintaining the home language to facilitate reintegration in case of remigration to the country of origin. Later, a number of aims were added or substituted, such as: the development of a positive self-image; offering possibilities for maintaining contacts with relatives and other family members living here or in the country of origin; reducing the gap between school and home environment; gaining access to one's own cultural heritage. As a language model, HLI can probably best be characterized as a restricted maintenance model.

There are several variants of HLI. In some cases a distinction is made between in- school and out-of-school HLI; in others between official and unofficial HLI. Currently, a large part of all non-indigenous primary school children have the possibility of getting five hours of official HLI a week, i.e., HLI subsidized and organized by the Dutch government. They are allowed to get no more than half of this during school time ("in school"), which amounts to half a school day. The other half they get after school hours or on their free Wednesday afternoons ("out of school"). Research (Inspectie van het

Home Language Proficiency 367

Onderwijs, 1988; Driessen, Jungbluth, & Louvenberg, 1988; Driessen, de Bot, & Jungbluth, 1989) shows that about 90% of all non-indigenous children receiving HLI do so during school time; on average the children receive about two hours of HLI a week. HLI provided outside school hours is mainly attended by the smaller language groups (e.g., Italians, Spaniards), here too, for an average of two hours a week. It is important to note that the Dutch government stipulates that the language taught during HLI should be the official language of the country, and therefore not another language or dialect spoken in that country (e.g., Kurdish for Turks or Berber for Moroccans). In addition to this official variant, there are also forms of HLI organized by the non-indigenous parents themselves or by embassies. In principle, there is no limit to the number of hours involved. Research carded out by Driessen et al. (1989) shows that 15% of the Turks and 30% of the Moroccans take part in this kind of out-of- school instruction, often alongside in-school HLI. In addition to these HLI variants, Moroccan and Turkish children very often also attend classes in the Koran, in which classical Arabic plays an important part. It is not exactly clear what is covered in these classes, and over what period of time. In principle, there is no limit to the number of hours pupils can receive this type of instruction; in some cases this is up to fifteen hours a week (Driessen, 1990; Wagenaar, 1993).

Right from the very start, HLI has faced a range of problems. The most recent studies into this subject (Driessen et al., 1988; Inspectie van het Onderwijs, 1988) summarize them as follows: There is a shortage of qualified teachers, their command of Dutch is often poor and their pedagogical-didactic approach is sometimes not in line with the one generally used in the Netherlands; there is a shortage of educational resources adapted to the Dutch situation as well as a shortage of classrooms; there are only occasional contacts and meetings with the Dutch team; there is disagreement about the aims to be achieved and the time that can be spent on these. The final conclusion is that HLI is still largely operating in complete isolation from regular Dutch education, even though the limiting conditions have improved over the years.

Finally, some figures to give an impression with regard to the scale of HLI: In 1991, a total of 66,000 pupils enrolled for official HLI. Of these, 60,000 belonged to the Turkish and Moroccan population, which constitutes about 75% of the total number of Turkish and Moroccan pupils. In 1991, official HLI took up a total o f f 63 million. 3

Research

Education involving home language has been an extremely controversial subject in the Netherlands right from the very start. Those involved can be roughly divided into two camps. On the one hand, there are the linguists who regard a certain command of the home language (i.e., the first language) as a necessary prerequisite for developing an adequate command of Dutch (in this case, the second language). 4 These linguists have, in addition, recently started to attribute an increasingly autonomous function to bilingual education (in this case, HLI), which involves learning the home language for its own sake. On the other hand, there are the educationalists and educational sociologists who do not attribute more than a very temporary supportive function to the home language in Dutch education. They would prefer to see a situation in which non-indigenous children start to learn Dutch immediately and do not have to bother about learning their home language. The disagreement between the two scientific

368 G. W. J. M. Driessen and M. A. van der Grinten

disciplines can, to a certain extent, be reduced to the old deficit versus difference controversy. 5 Advocates and opponents derive the evidence for their own particular points of view from the same (foreign) sources. On the basis of studies such as the ones described in the reviews by Baker & de Kanter (1983) and Willig (1985), linguists postulate that bilingual education most definitely is effective; educationalists use exactly the same studies to prove the contrary (see also Baker, 1987; Secada, 1987; Willig, 1987). This deadlock, which has existed since the very start of bilingual education, has not so far been broken (Lucassen & K~Sbben, 1992).

The occasion for the review at hand can be found in the recommendation, recently given to the government by the Committee for Non-indigenous Pupils in Education, about the future of HLI (Commissie Allochtone Leerlingen in het Onderwijs, 1992). 6 This recommendation makes out a case for drastic future expansion of HLI. In its recommendations the committee, however, largely ignores questions such as: What is already known about the effects on home-language proficiency? Is there at this scale a need for expansion, now and in the future? How long will this need continue to exist--if it exists at all--and, Can the expansion be realized in practice?

In the discussions concerning the role of the home language in Dutch education, attention is actually only paid to its use in the learning of Dutch. Very little is known about the effects it has on home language proficiency itself and this is something which is hardly ever brought up for discussion. 7 The review at hand hopes to provide a better insight into these effects. The amount of research carried out into in the field of home-language acquisition by non-indigenous people in the Netherlands, is in fact very limited. Three criteria were used for the final selection of the studies presently considered. (1) The studies are to pertain to Turks and Moroccans; there is hardly any material available on other language groups. (2) Furthermore, the studies must concern children in primary education; this type of education covers a period of eight years in total, the first two of which being nursery education and the last six actual "school years". From now on, we will be referring to these eight years as "forms". The children concerned are usually aged between four and twelve. (3) Finally, the studies should also establish a correlation between bilingual education and home-language proficiency. 8

The relevant studies are discussed in detail below. When results are summarized the emphasis is always on the level of home-language proficiency achieved. Subjects or relationships covered by the studies, but not directly relevant to the review (for instance, the correlation between the command of one's own language, on the one hand, and the command of Dutch or well-being at school, on the other) are therefore not discussed.

A fixed pattern is followed in describing the selected studies. The paragraph header refers to the researcher(s) involved in the evaluation. Subsequently the following issues are dealt with in succession: the educational model involved, the aim of the study; the design of the study; the sample; the research period, i.e., the data- collection period; the instruments used to establish the level of language proficiency (e.g., tests); the analysis techniques used; the most important research results; possibly some specific remarks that are of importance for the interpretation of the results; references which have not yet been quoted in the discussion, in order of importance.

Home Language Proficiency

Empirical Findings

369

Study 1: Teunissen

Model. An experimental bilingual model, in which in the third form during 55% of the time the Turkish and Moroccan children were taught in their home language and for the remaining 45% of the time in Dutch; in the fourth form, the ratios were reversed. Both years were spent in ethnically homogeneous classes. From the fifth form onwards, the children attended regular Dutch education (together with Dutch children) and in addition received 2.5 hours of HLI a week. A control group at other schools only received regular Dutch education and 2.5 hours of HLI a week.

Aim. A study into the effects of a bilingual-bicultural educational programme for Moroccan and Turkish pupils.

Design. Longitudinal, quasi-experimental, five cohorts. Sample. A local sample, in one city. Bilingual group: one primary school with 70

Turkish pupils and one primary school with 78 Moroccan pupils; control group: 20 primary schools with a total of 86 Turkish pupils. Age at the start: six to seven years old. When drawing samples, matches were made on starting level, age, participation in nursery education, social background, sex and home language.

Period. 1980-1984. Instruments. Two receptive oral tests: listening comprehension and vocabulary;

these were used to match the bilingual group with the control group. Three tests were used for the effect measurement itself: vocabulary, word decoding and reading comprehension. These were parallel tests, i.e., Turkish or Arabic translations of existing Dutch tests, or self-developed tests (derived from Turkish or Moroccan instructional materials). Reliability computations were not made for all of the tests; the number of items involved in each of the tests is not known either. Attention was paid to concurrent validity. By means of the latter three tests, the level of comprehension, verbal fluency and reading proficiency was tested in Turkish and Arabic, respectively; writing proficiency was not tested. Tests for matching were administered prior to the experiment (at the end of the second form), the tests for the effect measurement itself in the third and fourth form (i.e., during the experiment) and in the two subsequent forms, i.e., the fifth and sixth form. Not all of the tests were, however, administered to all of the pupils in all of the forms. One Arabic test (reading comprehension) could not be administered at all because it turned out to be too difficult. On average 30 to 40 pupils from each sub-group took part in the tests. Pupils repeating a year were tested a year later and their results were subsequently included in those of their original cohort.

Techniques. Table analysis, analysis of variance. Results. At the beginning of the experiment the vocabulary and listening

comprehension of the Turkish and Moroccan bilingual groups was no different from that of the Turkish control group. In the vocabulary, word decoding and reading comprehension tests the Turkish bilingual group scored significantly higher than the Turkish control group in all forms. The Moroccan bilingual group scored significantly lower than the Turkish bilingual group on all tests and in practically all forms.

As far as the development of language proficiency is concerned, the vocabulary test for the Turkish bilingual group showed an average change in score from 14.7 to 15.7 points (SD=2) and the Turkish control group from 12.3 to 12.5 (SD=3); the Moroccan bilingual group went from 11.9 to 14.5 points (SD-3). In the decoding test, the score for the Turkish bilingual group went from 19.4 to 62.7 (SD=I 1); for the Turkish control group scores went from 16.4 to 50.2 (SD---10) and for the Moroccan bilingual group from 19.8 to 24.0 (SD=6). It should be kept in mind that these development data pertain to a period of four years.


Teunissen concludes that the bilingual programme has a clear and fairly constant positive effect on the development of the native language of non-indigenous pupils, without seriously affecting their development in the Dutch language and related school subjects.

Remarks. For the Turkish bilingual model there was a control group; it was not possible, however, to find such a comparative group for the Moroccan bilingual model. Consequently, it is hard to interpret the test results of the Moroccan children from the experimental group. Teunissen compares the scores on the Turkish tests with those on the Arabic tests; it is very much the question whether it is possible to do this at all. Furthermore, it is not clear exactly which pupils were tested at which moments, with the aid of which instruments. To the extent that they can be interpreted, the test results show that the language development of the Moroccan pupils lags behind that of both the Turkish bilingual group and the Turkish control group. The positive effects noted by Teunissen, therefore only apply to Turkish bilingual classes.

References. Teunissen (1986a; 1986b).

Study 2: Verhoeven

Model. Two experimental bilingual models. The Turkish pupils in the third form of school A were first only taught in Turkish for a period of two months. After this, they attended lessons conducted in Turkish in separate mother tongue classes for about half of the time, while attending lessons in Dutch, together with Dutch pupils during the other half of the time. By the end of the fourth form, the Turkish and Dutch children were all receiving full-time education in Dutch; the Turkish children, however, had some extra hours of HLI. In the third form of school B the pupils were only taught in Turkish; it was not until the fourth form that instruction in Dutch was added. The model of school A is a variant of the simultaneous model, and that of school B of the transitional model. A control group of Turkish children also attended a few hours of HLI a week in addition to Dutch education, in the third and fourth form.

Aim. A study into the effect of two linguistic models on the acquisition of language proficiency in Turkish and Dutch.

Design. Longitudinal, quasi-experimental. Sample. Local sample, in four cities. Bilingual group: two primary schools, forms

3 and 4 (17 and 8 Turkish pupils, respectively); control group: ten primary schools, forms 3 and 4 (a total of 74 Turkish pupils). Average age at the start: 6.5. When drawing the sample, matches were made on such factors as home language, social background, participation in nursery education, age and the number of years that had to be repeated.

Period. 1982-1984. Instruments. Four tests to establish oral proficiency: phoneme discrimination (30

items), receptive vocabulary (108 items), productive vocabulary (80 items), sentence imitation (24 items). Three tests to establish reading proficiency: word recognition (reading out words correctly; 60 items), word spelling (32 items), reading comprehension (20 items). Parallel versions were developed for all of these tests: one in Dutch and one in Turkish. The tests were used to measure oral proficiency, reading and comprehension skills; writing proficiency was not included. The reliability (KR20, o~) of the tests varied from 0.83 to 0.96. Attention was paid to concurrent validity and content validity. The tests were administered three times: after one month, after ten months and after twenty months. However, as a result of a number of pupils dropping out of the experiment (because of remigratiori or having to repeat a year), only 70% of the pupils were tested at the last measurement.

Techniques. t-test, multi-variate analysis of variance, regression analysis, correlational analysis.


Results. At measuring moment 1 the oral proficiency levels of pupils in Turkish and Dutch were ascertained. Both the bilingual and the control group appeared to be significantly better in Turkish than in Dutch. As far as the development of oral Turkisfi proficiency during the three measuring moments is concerned, it became apparent that both groups showed significant progress at about equal rates; however, the bilingual group already started with higher scores at measuring moment 1.

With regard to reading proficiency, it was concluded that there were no significant differences between the experimental group and the control group in word recognition and word spelling; the pupils from the bilingual group were, however, better at Turkish and the pupils from the control group better at Dutch. As regards reading comprehension, there were no significant differences between the two groups, either for Turkish or for Dutch.

Verhoeven's final conclusion was that the education received by the pupils in the control group did not adequately link up with the background of the Turkish pupils, in view of the fact that for these pupils Turkish was still the dominant language even after two years of nursery education in Dutch. When comparing the two groups, it became apparent that the Turkish pupils in the bilingual model were achieving better results in Turkish and more or less comparable results in Dutch.

Remarks. Parallel versions were developed for each of the tests used; in general, this meant that the Dutch tests were "translated" into Turkish. Next, the results obtained in both tests were compared. This approach is based on the assumption that the underlying psycho-linguistic operations for both languages are identical, or at least comparable; it is questionable whether this assumption is in fact justified.

Due to pupils having to repeat a year or to remigration, both groups showed a drop-out rate of about 30% after one year. It is not clear how this affects the average level of the remaining pupils.

One remarkable conclusion (which was not drawn in the research report) is-- taking into account the differences in scores at the start of the research--that for the level of Turkish it apparently makes little or no difference whether pupils (in form 3) are taught in Turkish half of the time (the transitional model A), all of the time (simultaneous model B) or merely a few hours a week (the control group). In other words: the amount of instruction does not appear to affect the level achieved.

The research leaves a number of important questions unanswered, such as the one relating to the effect of the HLI received by the pupils in the control group, and another relating to the class sizes of the categories compared: If the two bilingual classes were (much) smaller than the control group classes--as the report implies--then it is possible that the effects are partly due to the fact that the children in the first group received much more individual attention from their teachers than the ones in the second group.

References. Verhoeven (1987).

Study 3: Driessen, de Bot and Jungbluth

Model. HLI. Practically all of the children received HLI, but the number of hours per week varied greatly.

Aim. A study into the correlation between, on the one hand, HLI participation, and on the other, knowledge of one's home language and culture, or performances in "regular" Dutch education.

Design. Cross-sectional, quasi-experimental. Sample. National, random sample of 120 primary schools containing all pupils in

the eighth form; 368 Turkish and 254 Moroccan pupils; average age: 12.5. Period. 1987/88.


Instruments. Written language tests (multiple choice and open questions) in Turkish (81 items) and Standard Arabic (53 items). As far as the linguistic description level was concerned, distinctions were made in the test between pragmatics, idiom, vocabulary, grammar and spelling. The level of the tests was adjusted to what the pupils should command, according to HLI teachers, policy makers, and educational and language experts. Pupil self-evaluation scales had scores ranging from (I) "I cannot do this" to (5) "I find this very easy" (12 items). Teacher's ratings featured scores ranging from (1) "definitely no command" to (5) "definite command" (four items). The reliability ((~) of the instruments ranged from 0.83 to 0.93. Face validity, concurrent validity and content validity were studied in more detail. The tests were used to measure reading and writing proficiency, the pupil self-evaluation scales and teacher's ratings were used to measure listening/comprehension, verbal fluency, reading and writing proficiency.

Techniques. Table analysis, correlational analysis, analysis of variance, regression and path analysis.

Results. The Turkish pupils got 73% of the test items right. Of the Moroccan pupils, 42% did not get a single item right; of the remaining 58 per cent, the average percentage of correctly answered items was 33. For Turkish and Moroccan pupils, the scores on the self-evaluation scale were 4.2 and 3.2 respectively. It should be noted that for the Moroccans verbal command was better than written command (3.5 and 3.0, respectively). On the teacher's rating scale the Turkish and Moroccan pupils scored 4.1 and 3.5, respectively.

In order to study the correlation between HLI participation and home language proficiency, HLI participation was indexed using the average number of hours per week during which the children had received HLI over the past years. Because of this, it was necessary to first of all check whether there were any differences between the various categories of HLI participation. The correlation between HLI participation and language proficiency was linear in only a few cases. For the Turkish pupils, the maximum nominal-metric correlation (Eta) with the three effect measures (test, scale and rating) was 0.27 and for the Moroccans it was 0.26.

The following characteristics appeared to be of (some) importance in explaining the differences in home language proficiency. For the Turkish pupils: HLI participation, age (negative), motivation, the importance which parents attach to HLI; the HLI teacher working towards the aim of "language maintenance". For the Moroccan pupils the most important factors were: language use at home (Moroccan Arabic versus Berber), the length of stay (negative), use of Dutch (negative), and the degree of contact between the HLI teacher and the parents of the pupils. Speaking a Berber dialect at home, in particular, had a strong negative effect on the command of Arabic. The linear correlation (r) between home language proficiency and Dutch language proficiency for Turkish pupils amounted to a maximum of 0.14 and for Moroccan pupils -0.22 (negative).

The final conclusion drawn by Driessen et al. is that there are major differences in language proficiency levels. The level of Turkish appears to be fairly good, whereas the level of spoken Arabic is poor and that of written Arabic definitely very low. The correlation between language level and HL1 attendance is weak to very weak.

Remarks. During the construction phase the Standard Arabic test had to be very much simplified, because a pilot study had shown that the pupils would not be able to attain the level originally aimed for; in addition, the test was produced in vocalized Arabic, which also meant a simplification.

It can be viewed as a disadvantage that one of the instruments used, the test, only measured general language proficiency, and was not sub-divided into sub-tests; this makes it hard to establish pupils' level on certain psycho-linguistic sub-aspects.

In a comparative study in three Moroccan cities involving 117 pupils (forms 2-4; aged between 8-12) the Standard Arabic test was administered again (Bentahila & Davies, 1990). It appeared that 47, 51 and 78% of the pupils in the second, third and


fourth form, respectively, were able to answer the test questions correctly. From this, Bentahila and Davies conclude, that the results in the Netherlands are not all that bad, compared with the results obtained in Morocco itself. In any case not as bad as they first appeared to be. According to Driessen & de Bot (1991), the findings in Morocco provide a double relativization for the low test results obtained in the Netherlands. On the one hand, the level aimed at in the Netherlands is far too high, and on the other hand, the level achieved in Morocco itself is very low.

References. Driessen et al. (1989), de Bot et al. (1991), Driessen (1990; 1991a; 1991b; 1992a; 1992b; 1993; 1994), Driessen and de Bot (1990), de Ruiter (1992).

Study 4: Van de Wetering

Model. HLI. The children received HLI but the extent to which they received it varied widely.

Aim. A study into the effectiveness of HLI to Moroccan children, and especially into the effects of education in the Arabic language.

Design. Longitudinal, quasi-experimental. Sample. Local sample, eight primary schools in two large cities; 447 Moroccan

pupils from forms 3-8; aged between 6-14. Period. 1983-1985. Instruments. Decoding test (63 words), reading comprehension test I (14 multiple

choice items), reading comprehension test II (13 multiple choice items). The decoding test measured reading proficiency and verbal fluency, and the reading comprehension tests measured the level of reading proficiency. Pupils were not required to produce any written work. All three tests were translated and edited versions of existing Dutch tests. The level of the tests was adjusted to the level at the end of third form of primary school. In this study, the reliability of the tests was not established, but reference was made to studies in which the reliability (o0 for reading comprehension was 0.63. Aspects of concurrent validity were taken up. The tests were administered three times, however, not always on the whole sample, since the experiment had a drop-out rate of over 45%.

Techniques. Table analysis. Results. The researcher presents the test results in relation to the number of years

of Arabic-language instruction the pupils had received, both in the Netherlands and in Morocco. In the word decoding test 71% of the pupils with three years of Arabic education or more got at least 33 of the items correct in the first research year; in the second research year, the same score was achieved by 72% of the pupils with four years of Arabic education or more, and in the third research year by 76% of the pupils with five years of Arabic education or more. In reading comprehension test I, 70% of the pupils with three years of Arabic education or more got at least 10 out of 14 questions right in the first year; in the second year the same score was achieved by 84% of the pupils with four years of Arabic education or more, and in the third year by 87% of the pupils with five years of Arabic education or more. In reading comprehension test II, 39% of the pupils got l0 or more of the 13 questions right, in the first year; the following year the same score was achieved by 33% of the pupils and in the last research year by 48% of the pupils.

Van de Wetering's final conclusion was that the yield of HLI in the field of word decoding and reading comprehension is relatively low; that at least three years are needed to round off the decoding process and that after five to six years of HLI pupils can be expected to voice and comprehend a simple vocalized text, on condition, however, that HLI functions under reasonably favourable circumstances.

Remarks. In Driessen (1991c) attention is drawn to the fact that no more than 54% of the pupils participated in the decoding test; for reading comprehension test I


this percentage was 24 and for reading comprehension test II it was a mere 9%. This was caused, on the one hand, by the fact that HLI teachers often did not allow their pupils to participate in the test administration because they felt their pupils could not meet the required level, anyway; and, on the other hand, by the fact that pupils who were not successful in the first test were not allowed to go on with the second or, respectively, third test. In addition, about half of the respondents dropped out in the course of the research. Furthermore, the number of test items was extremely small and only reading proficiency was tested; no research was carried out into writing proficiency. Consequently, the results presented undoubtedly very much over-estimate the level of the group as a whole. Driessen (1992c) therefore warns against drawing incorrect conclusions. He shows that the interpretation adopted by van de Wetering is not correct, or at least incomplete. From his re-analyses it appears that the level of language proficiency is extremely low and that this level furthermore goes down for each consecutive form as the years go by.

References. Van de Wetering (1988; 1990), Driessen (1994).

Study 5: De Ruiter, Aarts and Verhoeven

Model. HLI. The pupils participated in HLI to various degrees. Aim. The development of Turkish and Arabic tests which present a summarized

picture of the yield of HLI in primary education and also a diagnostic profile for HLI in secondary education.

Design. Cross-sectional, quasi-experimental. Sample. The study was made up of three parts: an exploratory study, the main

study (both in the Netherlands) and a comparative study (in Turkey or Morocco). The exploratory study pertains to a local sample: 23 primary schools in seven cities with 69 Turkish and 81 Moroccan pupils in the eighth form. For the main study, a national, random sample was drawn; 68 primary schools; 263 Turkish and 222 Moroccan pupils in the eighth form. The comparative study in Turkey pertains to a local sample survey: six primary schools in two cities, with 276 pupils in the highest form. The study in Morocco pertains to a local sample; six primary schools in two cities, with 242 pupils in the highest form.

Period. 1990, 1991. Instruments. Two tests to establish listening proficiency: vocabulary and

instruction comprehension; four test to establish reading proficiency: word decoding, spelling, vocabulary, syntax and reading comprehension. In the first instance a total of almost 300 multiple choice-items was used to determine receptive language proficiency, i.e., only understanding/comprehension and reading proficiency. In the exploratory study, two parallel versions were developed, i.e., a Turkish test and a Moroccan Arabic/Standard Arabic test. The reliability (ct) of the original tests for Arabic ranged from 0.38 to 0.93 and the reliability for Turkish tests from 0.57 to 0.92. After adjustment (by means of removing about 100 items), the reliability varied from 0.67 to 0.94. Part of the tests had been derived from Dutch material intended for six year olds. Content validity and concurrent validity were ascertained.

Techniques. t-test and correlational analysis. Results. Test scores were converted into the percentage of items answered

correctly. In the exploratory and the main study carried out in the Netherlands, the scores of the Turkish pupils varied between 52 and 96. A distinction can be made here between, on the one hand, listening proficiency, with scores between 75 and 87, and reading proficiency on the other, with scores between 52 and 67. The word-decoding test was not included in the latter; the scores on this were 93 and 96. In the comparative study carried out in Turkey, the scores were generally higher than those in the main study (by 5 to 18 test points), with the exception of those for reading


comprehension and instruction (6 points lower). The researchers concluded that in the Netherlands pupils did reasonably well in the spelling test and fairly satisfactorily in the reading comprehension test. On the other parts they achieved very reasonable to good scores. Furthermore, the researchers argued that the children in the Netherlands did not lag very far behind those in Turkey; the areas in which the pupils lagged behind most were spelling and written vocabulary.

For the Moroccan children the scores in the exploratory and the main study varied between 24 and 81. For listening proficiency, the scores were between 52 and 74 and for reading proficiency between 24 and 60. Here, again, the scores for word decoding were considerably higher, namely 81 and 79 respectively. In the comparative study in Morocco, the scores on all parts were considerably higher, i.e. 13 to 42 points compared to the main study. The conclusions were that listening proficiency was fair and that instruction comprehension was of a lower level. Word decoding presented few difficulties and the researchers also considered the level of reading comprehension to be satisfactory. However, the children did not have a command of spelling and syntax yet and their written vocabulary was not very extensive. According to the researchers, it was possible to conclude on the basis of the differences between the results obtained in the Netherlands and those obtained in Morocco that the children in the Netherlands have a better command then was found in earlier research (in this case, Driessen et al., 1989 and van de Wetering, 1990).

The final conclusion of the study is that the Turkish pupils in general have a high level of listening proficiency and reading proficiency in Turkish; the listening proficiency of the Moroccan children is reasonably good, but their reading proficiency is fairly poor.

In the exploratory study, the correlation between language proficiency and HLI participation was also examined. For the Turkish pupils (who had on average attended six years of HLI), the correlations (r) were between -0.01 and 0.31, of which the correlations for the oral tests were the highest. The relationships between the degree to which the children spoke Turkish at home and with friends (versus Dutch) were all weak to very weak, and sometimes even negative; the correlations varied from -0.18 to 0.28. Tests of the differences between first form entrants and higher form entrants (side streamers) showed that the group that had attended Dutch education from the first form onwards in general scored worse. For the Moroccan children the correlations between the test parts and HLI participation varied from -0.13 and 0.30. The correlation between the degree to which the pupils spoke Arabic at home or with friends (versus Dutch) and the test parts were also weak; they varied from -0.16 tot 0.38. In general, the differences in language proficiency between f'trst form entrants and higher form entrants were small and not significant at any point, which led the researchers to draw the conclusion that education attended in Morocco apparently does not have an effect on the test scores.

Remarks. Driessen's (1992d) criticism of this study is not just directed at its technical-methodological aspects, but also very much at the way in which the researchers interpret the results. His main objection is that the tests were limited to receptive language proficiency; the pupils were not expected to produce language. Therefore, the researchers only succeeded in finding out something about listening and reading skills (the easier skills), but nothing about verbal and writing skills (the more difficult skills). In view of this limitation, Driessen feels that the researchers very much overstepped their mark by concluding that the Turkish children had a "good" command of Turkish and the Moroccan children a "reasonably good" command of verbal Arabic and a "fairly poor" command of written Arabic. He states that if the children had been subjected to productive tests (as was the case in the research by Driessen et al., 1989) their level would no doubt have turned out to be considerably lower than the one suggested.

Another point of criticism raised (Driessen, 1993) relates to, among other things, the fact that on three out of the seven Moroccan tests the pupils did not, or hardly,


score above guessing level (four multiple choice items, with on average 24 to 33% of the items answered correctly). It is further stated that a very rosy picture was painted of the results of the exploratory study. From the research report (Aarssen, de Ruiter, & Verhoeven, 1992) it appears that only half of the pupils took part in some parts of the test. The remaining pupils were not allowed to participate in the tests by their HLI teachers, who felt that the tests were far too difficult and that their pupils would not be able to do them anyway. In other words: If these pupils were to have sat the tests, the general average would have dropped considerably as a result.

References. De Ruiter et al. (1992), Aarssen et al. (1993), de Ruiter (1992).

Study 6: Wagena~

Model. In an experimental bilingual model, in an ethnically homogeneous class, in the first and second form of primary education (nursery school) Moroccan Arabic was spoken in the mornings and Dutch in the afternoons (for respectively 15 and 8 hours per week). Originally the intention was to reverse the set up in the third form, but owing to the non-availability of Moroccan teachers it became necessary to switch to Dutch education with some additional hours of HLI per week after the second form. The third form did however remain ethnically homogeneous; in the fourth form the children were distributed over various ethnically heterogeneous classes. A control group consisted of Moroccan pupils receiving normal Dutch education as well as some hours of HLI a week.

Aim. A study into the effects of bilingual nursery education on the school career of Moroccan children.

Design. Longitudinal, quasi-experimental. Sample. A local sample, one primary school (bilingual model) and two primary

schools (control group) in a large city; two times 30 Moroccan children from the first, second and fourth form; two times 23 pupils who spoke Moroccan Arabic at home, and two times 7 pupils who spoke Berber. The average age at the start of the study was 4.5. In the sampling, matches were made against ethnic background, age, sex, and socio- economic background.

Period. 1987-1991. Instruments. The sub-tests for language comprehension and language

production (speaking) of the Reynell Developmental Language Test and the sub-tests for passive and active vocabulary of the Language Test for Non-indigenous Children. These tests were used to measure oral language proficiency; they were Moroccan Arabic translations of existing Dutch tests. No information is available about the content and number of items of the tests, and no data is presented about the reliability and validity of the tests. The tests were administered three times: at the beginning of the first form (pre-test), at the end of the second form (effect measurement), and at the end of the fourth form (post-test).

Techniques. Analysis of (co)variance, regression analysis, correlational analysis, discriminant analysis.

Results. The results of the first two measuring moments are expressed in age equivalents. In the first measurement, the bilingual group scored 3.8 years for Moroccan Arabic language comprehension and the control group scored 4.1 years (the norm is 4.5 years). There was only one significant effect for ethnic group: the scores of the Arabic-speaking children were 3.8 and 4.6 years, those of the Berber-speaking children were considerably lower, i.e., 2.5 and 2.4 years in the bilingual group, respectively, the control group. The language production level for Moroccan Arabic was much lower, for the bilingual and control group the scores were 2.8 and 2.9 years, which is more than 1.5 years below their norm. Here, too, ethnic group had a significant effect. The Arabic speaking children obtained scores of 2.9 and 3.2 years and the


Berber-speaking children obtained scores of 2.2 and 1.9 years. In all cases, the scores for Dutch language comprehension and language production were.about 2.5 years, which is below the level of Moroccan Arabic, and 2 years below the norm. According to Wagenaar, the latter provides an argument for offering bilingual education.

In the second measurement, which took place when the children were aged 6, the bilingual group's score for Moroccan Arabic language comprehension was 5.3 years and the control group's was 3.7 years; after checking for the first measurement this difference turned out to be a significant one. There was also a significant difference between Moroccan Arabic and Berber speaking children: 5.7 years and 4.5 years versus 4.9 and 2.9 years for the bilingual group and the control group, respectively. Therefore, the bilingual group was more than six months, and the control group two years, below the norm. In the bilingual group the Berber speaking children were more than one year below the norm. The picture for language production was comparable, the difference being that the overall level of verbal fluency was much lower than the level of comprehension. The scores for the bilingual group and the control group were 3.9 and 2.9 years, respectively. After checking the level for the first measurement this difference appeared to be a significant one. There were major differences between Moroccan Arabic speaking and Berber speaking children, but these were not significant. In the bilingual group both sub-groups appeared to have profited from the education, but the Arabic speaking children considerably more so than the Berber speaking ones. Wagenaar states that the level of the Berber children is still so low that one should have serious doubts about the use of Moroccan Arabic education for this group. She attributes this result to the fact that the home language of the Berber children is a Berber variant and not Moroccan Arabic.

Regression analysis was used to establish which factors could explain the differences in the level of Arabic. For language comprehension the ethnic origin (Moroccan Arabic versus Berber) appeared to be the most important predictive factor, followed by participation or non-participation in the experiment (fl=-0.48, andfl=0.28, respectively); the same factors applied for language production, although the strength of the coefficients differed (fl=-0.26 andfl=0.61, respectively). IQ, sex, educational level of parents and language orientation at home no longer appeared to have a significant effect.

In the third measurement, which took place two years after the experiment was completed, the expected norm for passive vocabulary was between 83 and 92. Although the scores of the bilingual group were considerably higher than those of the control group (62 versus 47), this difference appeared not to be significant. Both groups stayed well below the norm, despite the fact that both groups had attended HLI in the third and fourth form. The expected norm for active vocabulary was between 42 and 51; the scores of the bilingual group and control group were 18 and 17 respectively. After controlling for the starting measurement, this difference was not significant either. Wagenaar concluded that the home-language proficiency of the bilingual group continued to develop well during the period in which the experiment was being run, but stagnated when the experiment was stopped. Productive language proficiency in particular appears to be sensitive to education. According to the researcher, the Moroccan Arabic language is given insufficient support in the home environment to allow it to continue to develop.

A relatively large number of children from the bilingual group were referred to special education. Discriminant analysis showed that this was mainly related to Berber background and a very traditional family orientation on the one hand, and participation in the bilingual model, i.e., attending Moroccan Arabic education, on the other. On the basis of this Wagenaar advises Berber speaking children against participation in this type of bilingual nursery school.


Remarks. It is unfortunate that there are no norms available for the tests used, which were all Moroccan Arabic translations, when constant comparisons are being made with a norm group (based on the Dutch tests).

References. Wagenaar (1990; 1993), Wagenaar and Scholte (1989a; 1989b; 1990).

Conclusions

In the first place, it is remarkable that in 25 years of bilingual education no more than six studies have been financed in which the effect of this type of education on home-language proficiency is examined. In fact, only one study (by Driessen et al., 1989) into the effects of HLI was financed by the Ministry of Education and Science. 9 Its budget amounted to about 0.3 million guilders. This is in sharp contrast to the amount of money that has been invested in HLI over the last 25 years (roughly estimated to be around 1 billion guilders). The controversial and sensitive nature of this type of education could be seen as the cause for this. Just like in all other matters relating to ethnic minorities in general, politicians and policy makers are somewhat reluctant to make hard statements or take decisions with regard to this type of education (cf. Lucassen & Ktibben, 1992). They would rather steer clear of this type of issue than run the risk of getting their hands burnt. It is, after all, somewhat remarkable that there appears to be no more of a need to get some insight into the effects of an investment of this kind by the authorities (in financial terms) and by the children concerned (more than 10 per cent of the time they spend in primary education).

A second conclusion relates to the fact that part of the studies discussed are dated. The data for the studies by Verhoeven, by van de Wetering and Teunissen was collected more than ten years ago, the data for the studies by Driessen and Wagenaar more than five years ago. This is a serious shortcoming, particularly for the theme at hand, because there is every indication that the length of time the pupils have spent in the Netherlands--which increases over the years--is a very decisive factor for their command of the home language.

Another consideration relates to the size of the samples. Those of the studies involving transitional and simultaneous models (Teunissen, Verhoeven, Wagenaar) are particularly small, making generalization problematic.

An important issue is that of the home language itself. The bilingual programmes were first introduced in the Netherlands, because it was felt that non-indigenous pupils had the right to learn their home language (also referred to as "mother tongue," "native language" or "first language"), in order to be able to communicate with their family/relatives in that language, gain access to their own cultures, or acquire this language as an aid to facilitate the learning of Dutch. Only a few people have pointed out that for a very large part of the non-indigenous people concerned the official national language in which they are to be taught is not in fact their home language at all (Driessen, 1991a). This is also the case for the two groups discussed here. There is not a single Moroccan child that speaks Standard Arabic at home: they either speak a Moroccan Arabic dialect or a Berber variant (people speaking one dialect or variant can hardly, if at all, understand those speaking another dialect or variant, for that matter). To them, Standard Arabic is a completely foreign language. To many Turkish children, Standard Turkish is equally foreign. Turks originating from the rural areas,


especially, speak dialects that deviate from Standard Turkish to a greater or lesser extent. Furthermore, Turkey has some minority languages, of which Kurdish is probably the best-known one. For Kurds, in particular, who are oppressed in Turkey and whose language was officially forbidden until recently, it is psychologically unacceptable that they should learn Standard Turkish, the language of their oppressors, in bilingual education.

In view of this peculiar situation, some important aims of bilingual education definitely no longer hold any water. Command of the official national language, after all, is no guarantee that the children will be able to communicate with family members/relatives. Often the culture taught is not their "own" culture. Nor can it be expected that education in a language and culture which are not one's own will contribute to a positive self-image. Possible transfer-effects are also unlikely in this case: if a foreign standard language is taught instead of the mother tongue this in fact means that pupils have to learn a third, foreign language.

Another important point is the way in which the progress, the stagnation or deterioration in the language development of the individual or the groups is determined. This therefore involves the point of reference selected or the norm with which the test scores of pupils from the sample are to be compared (see Jaspaert, Kroon, & van Hout, 1986). One possibility is to compare form cohorts with varying lengths of stay or periods of instruction (see Driessen et al., 1989; van de Wetering, 1990) or to compare subsequent generations (e.g., de Ruiter, 1989; Huls & van de Mond, 1992). Another frequently used method is the one in which pupils in the immigration country are compared with pupils in the country of origin (see Driessen et al., 1989; de Ruiter, 1992; Boeschoten, 1990; Schaufeli, 1991). Usually, both groups are first matched on some background characteristics such as age, socio-economic background and language background, before the test results are compared. The (relativizing) conclusion which is then frequently drawn, is that the children in the immigration country (in this case, the Netherlands) do not lag very far behind those in the country of origin.

However, some words of caution are called for when drawing this type of conclusion. In the first place, it is possible to conclude from the study by Schaufeli (1991) that the tests administered in Turkey were biased. This bias was caused by the fact that the children in Turkey--as a result of the different pedagogic-didactic approach generally used there, which is based on competition and reproduction of knowledge--often were not accustomed to the type of tests they had to complete for the Dutch researchers. As a result the reliability and validity of the test results are questionable. At the same time this also means that the true level in Turkey might have been underestimated. This in turn could mean that the children in Turkey are in fact performing much better than the children in the Netherlands. Naturally, a similar conclusion would apply to Morocco. A second word of caution relates to the fact that these designs were not matched on one vital characteristic: education, especially the quality of that education. Characteristics such as organization, pedagogic-didactic approach and content were not included in the comparison, either. In other words, the fact that the level of education in the country of immigration and the country of origin can vary immensely is overlooked, as well as the fact that the level of language proficiency in the country of origin might have been considerably higher if this education had been of the same quality as that in the country of immigration.

380 G. W. J. M. Dries,sen and M. A. van der Grinten

The following example (partly) derived from the study by Schaufeli (1991) should illustrate this. This researcher compares Turkish children in the Netherlands with those in Turkey and points out the following problems: The children in Turkey often have to walk long distances in order to get to school; they are frequently absent, for instance, when the weather is bad or when they have to help out at home-- th is is particularly true for girls. As the children frequently have to help out at home, they do not have time to do homework for school. At home there is no tradition of attending school or reading, and children therefore do not receive any help with their homework, either. All the instructional materials, including the school books, have to be paid for by the (often needy) parents themselves; if there is no money, then the children do not have books. Children are taught in very heterogeneous classes with one (underpaid) teacher for (up to) 60 pupils. The pedagogic-didactic approach is characterized by discipline, imitation and role learning. When this is compared with the situation in the Netherlands (or any other western country) it is obvious that the level in the country of origin simply cannot appear anything but low. It is therefore unjustified to use the country of origin as a point of reference and subsequently be satisfied with the level achieved in the country of immigration. In view of the "luxurious" educational situation in the Netherlands, the level achieved here could be expected to be at least the same as that in the country of origin, despite the relatively limited period of instruction-- certainly when it comes to written language proficiency.

One last point refers to the language skills examined. A rough classification is one in which four language modalities are distinguished: understanding, speaking, reading and writing. The first and third are receptive skills and the second and fourth are productive skills. In general, it is assumed that the sequence in which the modalities are presented here is also one of increasing difficulty. Only one of the studies looked at in this review (Driessen et al., 1989) pays attention to the most difficult skill, writing (i.e., written language production); the other studies limit themselves to oral proficiency and reading. This is a serious limitation because they only shed light on a part of language proficiency. (The fact that practically all studies only make use of multiple-choice tests is a problem too.) This is most certainly something which will have to be taken into consideration in the interpretation of the research results.

By now, it will have become clear that in terms of methodology there are many flaws in the studies discussed. In this respect, the Dutch situation is no different from that existed in the USA ten years ago (Baker & de Kanter, 1983; Birman & Ginsberg, 1983; Willig, 1985). This means that it is almost impossible to draw hard conclusions about the effectiveness of bilingual programmes. Statements concerning the level of language skills as such, - - i.e., ones that do not take into consideration instructional characteristics - - are somewhat less risky to make but not entirely unproblematic either.

If we disregard this type of objection, we can point out the following trends in language skills as such. The level of Turkish appears to be reasonably good, both from an oral and a written point of view. As far as the level of Arabic is concerned, an emphatic distinction has to be made between Moroccan Arabic (the informal, spoken language) and Standard Arabic (the formal, written language). The command of the first is limited, that of the second downright poor. The difference in level can largely be attributed to the fact that to all Moroccans, Standard Arabic is a foreign language. Furthermore, the children do not live in Morocco but in the Netherlands, which makes it even more difficult for them to learn Standard Arabic: it is written in another direction

Home Language Proficiermy 381

(from the right to the left), uses different characters, and there is little communication in, and contact with, Standard Arabic.

The programmes' effect on the level of language proficiency is not entirely clear. Transitional and simultaneous models appear to be more effective than HLI programmes, which is probably related to the length of the period of instruction, the age at which it is provided, the length of stay and the home language of the children. As concerns the question about to the persistence of the effects of transitional and simultaneous models after completion of the experiments, the results are not promising (see Wagenaar, 1993). The correlation between HLI and home-language proficiency is at the very best weak, and at the very worst negative. For the language level, factors such as home language (e.g., Moroccan Arabic versus Berber or Dutch) and length of stay (the longer the period of stay in the Netherlands, the lower the level) appear to be at least as important as the (period of) instruction. Among Moroccan children in particular a very definite loss of language, or perhaps more aptly, a stagnation in language acquisition seems to take place. This is confirmed by linguistic studies carried out in the Netherlands, which have not been discussed here, such as those by Nortier (1989) and de Ruiter (1989) for Arabic, and Boeschoten (1990), Schaufeli (1991) and Huls and van de Mond (1992) for Turkish. One should probably not expect HLI to do a great deal more than simply slow this process down a little.

To return to the occasion for this review, i.e., the recommendation by the Committee for Non-indigenous Pupils in Education to drastically increase HLI, we have to conclude the following. In view of the fact that only very few evaluation studies have been carried out into the effects on the home language, and given that the little that is known about this matter has serious methodological shortcomings, there are, for the time being, no grounds for deciding whether this increase is justified or not. l0 One should really first of all carry out some adequate research, i.e., a longitudinal study, with adequate verification for the characteristics of the starting situation, with sufficiently large groups of pupils, with verification of the (educational) activities undertaken within the framework of the intervention (or in this case, the programme), and with tests which differentiate adequately according to the various language modalities and psycho-linguistic sub-skills. Above all, however, political will and courage are needed to bring the issue up for discussion and subsequently support its evaluation. At the same time, a needs analysis and a feasibility study should be carded out, too. That is: one should find out from non-indigenous parents and their children whether they have a need (now and in the future) for education in their home language. At the same time, they should be asked in which language this should take place, the actual home language (the mother tongue) or the official national language (the standard language). l l Linguists, educational experts and developers of language methods should then find out whether the wishes of the non-indigenous people can be realized at all. For instance, is it possible to develop an adequate teaching method for a non- codified language such as Berber and what kind of aims and norms should one use for this? One almost insurmountable problem in all this, is, however, the fact that time continues to pass, and the situation therefore continues to change. It is for this reason very debatable whether the results of such studies will still be relevant to the situation, by the time they become available.


Notes

1.

2.

3. 4.

5.

6.

7.

8.

9.

10.

11.

The number of non-indigenous people in the various statistics can vary greatly, depending on the criteria used. Here we started out from the country of birth of the person or his/her parents.

The Dutch name for this type of education is "Onderwijs in Eigen Taal en Cul tuur"--OETC (Minori ty Language and Culture Teaching). Originally, it consisted of a language component as well as a cultural component. The latter component was dropped, however, in 1991. Since then it has therefore been regarded as language instruction. f 1.00 NL = $ 0.54 = £ 0.36. They were particularly inspired in this by the interdependence and threshold hypothesis of Cummins (1979).

In short, the deficit view assumes that there are language delays among (in this case) Turkish and Moroccan children, and that these children could possibly catch up by means of compensation programmes. In the difference view all languages are considered to be equal---one can therefore at the very most speak in terms of language differences. Typical for this deadlock is that a year later another advisory body, i.e., the Social and Cultural Planning Bureau, on the basis of a study of the literature on the effects of HLI on Dutch educational performance, recommended that HLI should be scrapped from the primary-school curriculum entirely, and that in view of the considerable educational disadvantages of these children, all attention should be focused on the core subjects of Dutch language and arithmetic. They also recommended that if non-indigenous parents so wished, HLI could be organized outside school hours and that this type of education should only be subsidized by the Dutch government as long as it met certain quality requirements (Tesser, 1993). The review by Baker and de Kanter (1983) with regard to the situation in the U.S. did not go into the effects of bilingual education on performance in the home language, either. This means that research into bilingual projects in which only the level of Dutch language proficiency is tested (Appel, 1984) has not been included in this review. The study by van de Gelder (1982) relating to Spanish bilingual education has not been included, either. Monolingual studies into the home language proficiency of Turks and Moroccans (Nortier, 1989; de Ruiter, 1989; Boeschoten, 1990; Schaufeli, 1991; Huls & van de Mond, 1992) in which no link is made with education in the languages concerned, or in which adults form the target group, are not covered in this review. The project by de Ruiter et al. (1992) was related to the development of tests as such. The other studies were financed by scientific bodies or municipalities. According to some (see Tesser, 1993) it is not just the question of the possible expansion of HLI that needs to be raised, but also HLrs entire right to exist. Furthermore, in trying to answer this question one should also look at other effects, for instance, the effect on achievement levels in Dutch education. A poll carried out in 1987/88 into the need for education in non-standard languages in secondary education (de Jong, Mol & Oirbans, 1988) showed that there was only a very limited need for this. The most important consideration in the choice of the language was not so much its function, but its status. The researchers warn that any introduction of education in non-standard languages could encounter major difficulties, not just because of a lack of teachers and educational materials, but even more so because most parents would not be willing to allow their children to participate in this type of education.


References

Aarssen, J., de Ruiter, J.J., & Verhoeven, L. (1992). Toetsing Turks en Arabisch aan het einde van het basisonderwijs [Testing Turkish and Arabic at the end of primary education]. Tilburg: Tilburg University Press.

Aarssen, J., de Ruiter, J.J., & Verhoeven, L. (1993). Summative assessment of ethnic group language proficiency. In G. Extra & L. Verhoeven (Eds), Immigrant languages in Europe (pp. 159- 179). Clevedon: Multilingual Matters.

Appel, R. (1984). Immigrant children learning Dutch. Socio-linguistic and psycho-linguistic aspects of second language acquisition. Dordrecht: Foils.

Baker, K. (1987). Comment on Willig's "A recta-analysis of selected studies in he effectiveness of bilingual education." Review of Educational Research, 57, 351-362.

Baker, K., & de Kanter, A. (1983). Federal policy and the effectiveness of bilingual education. In K. Baker & A. de Kanter (Eds.), Bilingual education. A reappraisal of federal policy (pp. 33- 86). Lexington: Lexington Books.

Bentahila, A., & Davies, E. (1990). On the evaluation of Arabic language and culture lessons: Another perspective. Fez: University of Fez.

Birman, B., & Ginsberg, A. (1983). Introduction: Addressing needs of language-minority children. In K. Baker & A. de Kanter (Eds), Bilingual education. A reappraisal of federal policy (pp. ix-xxi). Lexington: Lexington Books.

Boeschoten, H. (1990). Acquisition of Turkish by immigrant children. A multiple case study of Turkish children in the Netherlands aged 4 to 6. Tilburg: KUB.

de Bot, K., Driessen, G., & Jungbluth, P. (1991). An evaluation of migrant language teaching in the Netherlands. In K. Jaspaert & S. Kroon (Eds.), Ethnic minority languages in education (pp. 25-35). Amsterdam/Lisse/Berwyn PA: Swets & Zeitlinger.

Commissie Allochtone Leerlingen in het Onderwijs (1992). Ceders in de tuin. Naar een nieuwe opzet van het onderwijsbeleid voor allochtone leerlingen [New educational policies for non- indigenous pupils]. 's-Gravenhage: DOP.

Cummins, J. (1979). Linguistic interdependence and the educational development of bilingual children. Review of Educational Research, 49, 222-251.

Driessen, G. (1990). De onderwijspositie van allochtone leerlingen. De rol van sociaal-economische en etnisch-culturele factoren, met speciale aandacht voor het Onderwijs in Eigen Taal en Cultuur [The educational attainments of non-indigenous pupils. The role of socio-economic and ethnic-cultural factors, with special attention to the Teaching of Migrant Language and Culture]. Nijmegen: ITS.

Driessen, G. (1991a). Landstaal of moedenahal? Het problematische karakter van de "eigen taal" binnen het Marokkaanse OET(C) [Standard language or home language? The problematic character of the "mother tongue" in Moroccan mother tongue teaching]. Migrantenstudies, 7 (2), 2-14.

384 G. W. d. M Driessen and M. A. van der Grinten

Driessen, G. (1991b). Marokkaanse kinderen op Nederlandse scholen. Een exploratie van hun achtergronden en onderwijsprestaties [Moroccan children in Dutch schools. An exploration of their backgrounds and educational attainments]. Tijdschri f t voor Onderwijswetenschappen, 21, 292-306.

Driessen, G. (1991c). Review of W. van de Wetering, "Onderwijs in Eigen Taal en Cultuur aan Marokkaanse kinderen in Nederland. Het OETC als resultante van een maatschappelijk krachtenspel'. Migrantenstudies, 7, (4), 55-57.

Driessen, G. (1992a). First and second language proficiency: prospects for Turkish and Moroccan children in the Netherlands. Language, Culture and Curriculum, 5, 23-40.

Driessen, G. (1992b). Ontwikkelingen in T1- en T2-vaardigheidsniveau van Turkse en Marokkaanse leerlingen [Developments in L! and L2 proficiency of Turkish and Moroccan pupils]. Toegepaste Taalwetenschap in Artikelen, 41, 27-35,76.

Driessen, G. (1992c). Her niveau Arabisch van Marokkaanse kinderen. Een waarschuwing voor misplaatst optimisme [Proficiency in Arabic of Moroccan children. A warning against false optimism]. Stimulans, 10, (5), 4-6.

Driessen, G. (1992d). Reactie op toetsing. Receptieve maar ook produktieve taalvaardigheid [A reaction on testing.- Both receptive and productive language skills]. Samenwijs, 12, 475.

Driessen, G. (1993). Toetsing Arabisch. Over realiteitszin en aanverwante zaken [Testing Arabic. On sense of reality]. Samenwijs, 13, 234-236.

Driessen, G. (1994). Moroccan children acquiring Arabic in the Netherlands. Mother tongue instruction as a means of combating educational disadvantage? In G. Driessen & P. Jungbluth (Eds), Educational opportunities. Tackling ethnic, class and gender inequality through research (pp. 71-88). Miinster/New York: Waxmann.

Driessen, G., & de Bot, K. (1990). Turkse en Nederlandse taalvaardigheid en leerlingkenmerken [Turkish and Dutch language proficiency and pupil characteristics]. Toegepaste Taalwetenschap in Artikelen, 37, 83-91,137.

Driessen, G., & de Bot, K. (1991) De eigen-taalvaardigheid van Marokkaanse leerlingen. Een nadere relativering [Home language proficiency of Moroccan pupils. A relativisation]. Levende Talen, 459, 124-127.

Driessen, G., Jungbluth, P., & Louvenberg, J. (1988). OETC in het basisonderwijs. Doelopvattingen, leerkrachten, leermiddelen en omvang [The Teaching of Migrant Language and Culture in primary education. Aims, teachers, materials and extent]. 's-Gravenhage: SVO.

Driessen, G., de Bot, K., & Jungbluth, P. (1989). De effectiviteit van her Onderwijs in Eigen Taal en Cultuur. Prestaties van Marokkaanse, Spaanse en Turkse leerlingen [Effects of the Teaching of Migrant Language and Culture. Attainments of Moroccan, Spanish and Turkish pupils]. Nijmegen: ITS.

de Gelder, A. (1982). Onderwijs aan Spaanse kinderen. Verslag van een onderzoek [Spanish children in education]. Hengelo: Stichting Buitenlandse Werknemers Twente.


Huls, E., & van de Mond, A. (1992). Some aspects of language attrition in Turkish families in the Netherlands. In W. Fase, K. Jaspaert & S. Kroon (Eds.), Maintenance and loss of minority languages (pp. 99-115). Amsterdam/Philadelphia: John Benjamins.

Inspectie van het Onderwijs (1988). OETC: niet apart maar samen. Een verkennend onderzoek naar de stand van zaken van her Onderwijs in Eigen Taal en Cultuur in het basisonderwijs [The Teaching of Migrant Language and Culture in primary education]. 's-Gravenhage: Mirtisterie van Onderwijs en Wetenschappen.

Jaspaert, K., Kroon, S., & van Hout, R. (1986). Points of reference in first-language loss research. In B. Weltens, K. de Bot & T. van Els (Eds.), Language attrition in progress (pp. 37-49). Dordrecht: Foris.

de Jong, M.J., Mol, A., & Oirbans, P. (1988). Zoveel talen, zoveel zinnen. De behoefte aan lessen Eigen Taal in het V.O. [The need for Home Language Instruction in secondary education]. Rotterdam: EUR.

Lucassen, L., & K~Sbben, A. (1992). Her parti~le gelijk. Controverses over het onderwijs in de eigen taal en cultuur en de rol daarbij van beleid en wetenschap (1951-1991) [The partial right. Controversies on the Teaching of Migrant Language and Culture and the role of government policy and science]. Amsterdam/Lisse: Swets & Zeitlinger.

Nortier, J. (1989). Dutch and Moroccan Arabic in contact. Code switching among Moroccans in the Netherlands. Amsterdam: UvA.

de Ruiter, J.J. (1989). Young Moroccans in the Netherlands: An integral approach to their language situation and acquisition of Dutch. Utrecht: RUU.

de Ruiter, J.J. (1992). Toetsing Arabisch: een vergelijking [Testing Arabic. A comparison]. Samenwijs, 13, 139.

de Ruiter, J.J., Aarts, R., & Verhoeven, L. (1992). Kennis eigen taal onderzocht. Toetsing beheersing Turkse en Arabische taal aan het einde van de basisschooi [Home language skills. Testing Turkish and Arabic at the end of primary education]. Samenwijs, 12, 395-399.

Schaufeli, A. (1991). Turkish in an immigrant setting. A comparative study of the first language of monolingual and bilingual Turkish children. Amsterdam: UvA.

Secada, W. (1987). This is 1987, not 1980. A comment on a comment. Review of Educational Research, 57, 377-384.

Tesser, P. (1993). Rapportage minderheden 1993 [Report on ethnic minorities 1993]. Rijswijk: SCP.

Teunissen, J. (1986a). Een school, twee talen. Een onderzoek naar de effecten van een tweetalig- bicultureel onderwijsprogramma voor Marokkaanse en Turkse leerlingen op basisscholen te Enschede [One school, two languages. Effects of a bilingual-bicultural educational programme for Moroccan and Turkish pupils in primary schools]. Utrecht: RUU.

Teunissen, F. (1986b). Project moedertaalklassen. Tweetalig en bicultureel onderwijs voor de tweede generatie [Mother tongue classes. Bilingual and bicultural education for the second generation]. Moer, 1986, 149-156.

Verhoeven, L. (1987). Ethnic minority children acquiring literacy. Tilburg: KUB.

386 G. W. J. M Driessen and M. A. van der Grinten

Wagenaar, E. (1990). Kleuterklas en moedertaal. Een onderzoek naar de effecten van tweetalig onderwijs aan Marokkaanse kleuters [Nursery class and mother tongue. A study into the effects of biligual education to Moroccan .children in nursery school]. Delft: Eburon.

Wagenaar, E. (1993). Tweetaligheid in het aanvangsonderwijs. Een onderzoek naar de effecten van tweetalig kleuteronderwijs op de schoolloopbaan van Marokkaanse kinderen [Effects of a biligual programme for Moroccan children in nursery school]. Amsterdam: Het Spinhuis.

Wagenaar, E., & Scholte, E. (1989a). Tweetalig onderwijs aan jonge Marokkaanse kinderen [Bilingual education for young Moroccan children]. Pedagogisch Tijdschrif~, 14, 231-244.

Wagenaar, E., & Scholte, E. (1989b). Effecten van tweetalig onderwijs aan Marokkaanse kleuters [Effects of bilingual education for Moroccan infants]. Tijdschrift voor Onderwijsresearch, 14, 227-236.

Wagenaar, E., & Scholte, E. (1990). Tweetalig onderwijs voor jonge Marokkaanse kinderen in het basisonderwijs: een replicatie-onderzoek [Bilingual education for young Moroccan children in primary education: a replication study]. Pedagogisch Tijdschrift, 15, 315-323.

van de Wetering, S. (1988). Her Onderwijs in Eigen Taal en Cultuur aan Marokkaanse kinderen in Nederland. lnterimverslag 3: Effecten van bet onderwijs in de Arabische taal [Mother Tongue and Culture Teaching for Moroccan children in the Netherlands. Effects of Arabic language teaching]. Utrecht: RUU.

van de Wetering, W. (1990), Onderwijs in Eigen Taal en Cultuur aan Marokkaanse kinderen in Nederland. Het OETC als resultante van een maatschappelijk krachtenspel [Mother Tongue and Culture Teaching for Moroccan children in the Netherlands. Arabic language and culture teaching as a result of social forces]. Utrecht: RUU.

Willig, A. (1985). A meta-analysis of selected studies on the effectiveness of bilingual education. Review of Educational Research, 55, 269-317.

Willig, A. (1987). Examining biligual education research through meta-analysis and narrative review: a response to Baker. Review of Educational Research, 57, 363-376.

The Authors

G E E R T W.J .M. D R I E S S E N and M I C H I E L A. V A N DER G R I N T E N are educa t iona l r esea rchers at the Inst i tute for A p p l i e d Soc ia l Sc iences ( ITS) at the Un ive r s i ty o f Ni jmegen, the Netherlands. Their major research interests are first and second language acqu i s i t ion , the t each ing o f minor i ty l anguage and cul ture , the pos i t i on o f non- indigenous pupi ls in educat ion, and educat ional careers.