Swedish ” vs . “ Non-Swedish ” . Immigrant Background and Cross-linguistic Influence in the Learning of English as a Foreign Language

<jats:p>-</jats:p>


Introduction
In an increasingly globalized world, command of foreign languages -English in particular -is widely regarded as essential to any country wishing to retain, or boost, its competitive edge in international affairs, whether in business or culture.Accordingly, huge amounts of money are spent worldwide on foreign language education.Not least in Europe, cooperation in this field between educational authorities in different countries has been steadily increasing since the 1980s, picking up speed after the collapse of the Soviet Union in 1989 and the subsequent expansion of the European Union.To European citizens at large, the individual benefits of such cooperation, aimed at improving language teaching, should be obvious.In a wider politico-cultural context, it may even be argued that "attainment of foreign language skills has a major part to play in the construction of Europe" (Bonnet 2004:165).
In view of the long tradition of foreign language teaching in Europe (cf.Kelly 1969), as well as the financial resources currently allocated to it, the idea of systematic intercountry comparisons of achievements in and attitudes towards foreign language instruction is hardly far-fetched.For instance, given the discrepancies with regard to national educational traditions, school systems, views on language teaching/learning, etc., what differences are there in listening and reading comprehension, oral and written proficiency, free written production, accuracy/correctness, etc.? In this article, a recent comparative study dealing with such issues provides the data investigated.
In 2002, a survey of acquired skills in English at the end of compulsory school (grade 9) was carried out in the following eight countries: Denmark, Finland, France, Germany, The Netherlands, Norway, Spain, and Sweden.This project, The Assessment of Pupils' Skills in English in Eight European Countries 2002 (henceforth Assessment of English;see Bonnet 2004, Skolverket/Erickson 2004), was an expanded follow-up to a similar survey of students' attainments in English as a foreign language conducted in 1996, involving Finland, France, Portugal, Spain, and Sweden (Bonnet 2004:11, Skolverket/Erickson 2004:8, Erickson 2005:45).The reason why English -rather than some other language(s) -was selected for the original study, and for the follow-up, was simply that English occupies a prominent place in foreign language teaching provision, not just because of the number of pupils learning it but also because of its status in the world.It is on the basis of this practical consideration that English was singled out.As all European countries teach English in their schools on a large scale they all have it as common ground for comparison and reflection on language teaching.No other European language offers this opportunity.(Bonnet 2004:11) Outlining the background and aims of the 2002 project, Bonnet notes (p.12) that, despite the vast amounts of taxpayers' money going into the teaching of English as a foreign language in the European Union, "the fundamental question of how successful this effort turns out to be is hardly ever formulated on a national -let alone European -basis."Consequently, "the aim ... is to set the ball rolling by making available a modicum of information on this central issue ...
[providing] some further comparative data on pupils' achievements across countries with a view to underpinning the cultural, structural and technical reasons why outcomes vary from one country to the next" (p.12).As in the 1996 survey, the three language skills tested were listening, reading and writing; "speaking was left out for both methodological and financial reasons" (p.13).In connection with each group of test items, students were asked to give their opinion as to their perceived degree of difficulty.There was also a self-assessment instrument ("can-do statements") concerning the skills tested.More specific information about the test takers and their teachers (personal background, attitudes to English, etc.) was provided by the inclusion of student and teacher questionnaires.Bonnet (2004:13) plays down the comparability aspect of the study, pointing out that "the very philosophy of the present approach is to provide broad indications about pupils' performance rather than attempt comparisons to the nearest decimal."He is careful to stress that the basic intention should not be seen as "an attempt to benchmark countries", and that " [t]here are perfectly well understood reasons why some countries' pupils perform better in English than others"; in fact, "the emphasis of this study is as much about what goes on in countries as about comparisons between them."The last point is of particular relevance for the purposes of the present study.
In Sweden, 1431 grade-9 students took part in the Assessment of English survey.Their results -which (together with those of Norway) were the highest among the eight countries participating in the study (Skolverket/Erickson 2004:27) -are summarized as follows by Erickson & Lagergren (2004:155): Overall the Swedish students performed well on the tests, especially on the parts focusing on receptive skills.… However, the distribution of results both within and between classes is considerable.There is a small group of students with very poor results, especially on tasks requiring some kind of written production.
Erickson & Lagergren note that "[n]ine out of ten students were born in Sweden …, and roughly the same number use mainly Swedish in their homes" (p.155).Further, among the "small group of distinct low-achievers" in the study, "[t]he proportion of students born in another country than Sweden and/or mostly speaking another language than Swedish at home is somewhat larger … than in the sample as a whole, both in the present study and in the national tests" (p.158).At the same time, the difference between students with an immigrant background, seen as a group, and the total Swedish result is quite small (Skolverket /Erickson 2004:65).
However, it would be grossly misleading to view the ten per cent of students with an immigrant or "foreign" background as a homogeneous group, "foreign" here being defined as born outside Sweden and/or speaking another language than Swedish at home.In fact, as emphasized by Skolverket/ Erickson (2004:63), these students differ along several dimensions, such as country of birth, language mainly used at home (henceforth "home language"), structural distance between English and home language, etc. -not to mention other relevant aspects for which information is not available, such as age of coming to Sweden for those not born in the country.
It would seem, then, that the "foreign-background" group of students in the Assessment of English survey deserves closer scrutiny, for several reasons and along different dimensions.In this study, I shall pay special attention to those students who, regardless of country of birth, have stated another language than Swedish as their home language.This subgroup will be referred to as the "non-Swedish" group -altogether 142 students (see further section 3) -which is thus more restricted than the "foreign-background" group as a whole.My basic aim in this study, which is not of a strictly quantitative nature in the sense of laying claim to statistically significant results (cf.note 9 below), is to throw some further light on the achievements of these students in the area of written proficiency, i.e. those test items in Assessment of English that deal with writing English (in a very narrow sense; see section 2).In particular, I shall focus on the following, partly intersecting issues in relation to the results of the "non-Swedish" group: overall difference between the "all-Swedish" group (born in Sweden, home language: Swedish) and the "non-Swedish" group; differences between groups of "non-Swedish" students with different home languages (e.g.Arabic, Spanish or Somali); differences within the "non-Swedish" group between those born in Sweden and those born elsewhere.
A further -more elusive -linguistic aim is to see to what extent it is possible to determine the role played by cross-linguistic influence ("transfer"; cf.section 5.1) in the learning of a specific field of English grammar, viz. the indefinite and definite articles.This discussion -of a mainly qualitative kind, although based on observed differences in test results between different home-language groups in the material studied -involves the relative typological ("contrastive") distance, with regard to a grammatical subsystem, between students' home languages and English, i.e. the likelihood of transfer, positive or negative, from different first languages (L1), but also from Swedish as a second language (L2) (cf.e.g.Ringbom 1987Ringbom , 2007;;Odlin 1989Odlin , 2003;;Ohlander 2001;Swan & Smith 2001;Jarvis & Pavlenko 2008).Such reasoning will here concern specific items in the Assessment of English test, relating to the English articles.For example, in the test sentence They are fond of ___ science fiction films, a number of students having Arabic as their home language chose to insert the indefinite singular article a/an (instead of "zero") in front of the plural noun phrase.To what extent, and with what degree of certainty, could this be attributed to cross-linguistic influence from Arabic?

Material
The following condensed account of the material used in this study is chiefly based on Bonnet (2004:15f.)and Skolverket/ Erickson (2004:18ff.).The total test battery of Assessment of English consisted of ten "exercises" (i.e.groups of items, henceforth "ex."), making up 79 points: listening comprehension (ex.1-3: 17 points), linguistic competence in the narrow sense of correctness (written; ex.4-7: 25 points), reading comprehension (ex.8: 16 points), written production (ex.9-10: 21 points). 2 Due to various problems of marking, certain items were later excluded, resulting in an aggregate score of 65 points instead of 79 (Skolverket/ Erickson 2004:26, 30f.).For the purposes of this paper, however, the original 79 points will be retained as a rough and ready basis for comparison between the overall results of different groups of test takers (section 4).In this connection, Erickson & Lagergren (2004:157) emphasize that "the tasks used in the present study cover only limited domains of foreign language proficiency." More specifically, the material focused on here is -with one exception -made up of the answers to the exercises designed to test linguistic competence and written production in a narrow sense (see below).Since ex.6 (5 points), dealing 2 With regard to the test battery, "it was decided that the test used in the 1996 survey should be kept, partly for practical and financial reasons, partly because it would enable longitudinal comparisons in the countries taking part for the second time.... The assessment tasks were taken from French and Swedish national tests of English from the mid-1990s" (Erickson 2005:46;cf. Bonnet 2004:165).For a comparison between the 1996 and 2002 surveys, see Skolverket/ Erickson (2004:80ff.).For principles and details concerning assessment and marking, see Bonnet 2004:68 (e.g., "missing data were re-coded as wrong"); cf. also Skolverket/Erickson (2004:20, 26).
with conversion of direct speech to indirect, gave rise to extensive marking problems, it was excluded from the assessment test as well as from the present study.This left ex.4, 5, 7, 9, and 10, comprising altogether 41 points, as my point of departure with regard to linguistic competence and written production.
The skeleton outline provided below of these five exercises will give a general idea of their nature and format (for the complete test, see Bonnet 2004:165ff.; Erickson/Skolverket 2004:18f., 96ff.).Although instructions to test takers in Sweden were in Swedish, they are here given in their English version (Bonnet 2004:169ff.).
Ex. 4 (4 points): quantitative determiners (some, any, etc.); four separate sentences, one gap in each, four alternatives; "Circle the alternative that makes the sentence correct"; e.g., There isn't ___ milk left in the fridge, alternatives: many, every, a little, any.Ex. 5 (4 points): articles; four separate sentences, one gap in each, four alternatives; "Circle the alternative that makes the sentence correct"; e.g., I've bought you ___ interesting book, alternatives: a, "zero", an, some.Ex 7 (12 points): tenses, verb forms; continuous text (story), twelve gaps; "First read the whole text and then fill in the correct form of each verb in brackets"; e.g., Do you know what happened to me yesterday?A week ago, when I (come) ___ home in the evening, I (find) ___ a little note on the kitchen table saying: … Ex. 9 (5 points): verb forms, do-construction; dialogue text (discussion about a film); "Fill in what is missing in the dialogue below"; e.g., Jean: I went to the cinema yesterday.Robert: ___ ?Jean: I saw Mrs Doubtfire.Ex. 10 (16 points): vocabulary/phraseology, prepositions, etc.; continuous text ("Working in a family"); cloze format ("rational deletion"), "Read the text in which there are gaps for words that are missing.Fill in the word that you think best fits in the context"; e.g., The chance of working as an au pair has been welcomed ___ thousands of young people.They come from many different ___ outside Britain, for instance Holland, Sweden and France.
As can be seen, the above exercises, under the headings (as specified in the test itself) of "linguistic competencies" (ex. 4-5, 7: 20 points) and "written production" (ex.9-10: 21 points), involve multiple choice (selected response: four alternatives) or gaps (constructed response).In other words, as stressed by Erickson & Lagergren (2004:157), the element of writing is severely restricted: "productive competence is hardly assessed …: writing is tested only in a very narrow way, e.g. in a cloze test", thus a far cry from "free written production" (cf.Köhlmyr 2003).In addition, ex.9-10 would seem to be as much about reading comprehension as about written production (cf.Skolverket /Erickson 2004:34).Obviously, then, notions like "linguistic competence" and "written production" as used in Assessment of English are considerably narrower than is usually the case; the test items accounted for above are not ideal for assessing students' general levels of "linguistic competence" or "free written production".This should be kept in mind throughout this paper. 3onetheless, despite the design weaknesses of the test as a whole and of certain test items, the results of the whole Assessment of English survey, as well as those of specific domains (e.g.receptive skills, writing), hold a great deal of interest.This applies to overall comparisons between countries, but also to comparisons between different domains; of no less interest are comparisons between subgroups within the same country.
3 The "non-Swedish" group: other home languages than Swedish As mentioned earlier, Norway and Sweden attained the highest overall results on the Assessment of English test, the Swedish group -1431 students -performing especially well on tasks focused on receptive skills.In linguistic competence/ correctness and written production, the Swedish results are somewhat weaker, although by no means unsatisfactory in relation to official national goals (Erickson & Lagergren 2004:155, 157;Skolverket/Erickson 2004:33, 36f.). 4 The overall picture further indicates that, in comparison with other countries, "the distribution of results in Sweden is not very wide" and there are "only minor gender differences" (Erickson & Lagergren 2004:158). 5f special interest here is that, as mentioned above, the overall result of students with an "immigrant" or "foreign" background" (born outside Sweden and/or with another home language than Swedish) is lower than the total Swedish result, but only to a very modest degree; in fact, the result of this subgroup is roughly on a par with the overall results of Denmark, Finland and The Netherlands, and markedly better than those of France and Spain (Skolverket /Erickson 2004:65).Thus, from an international perspective, the "foreign" students as a group do quite well on the Assessment of English test.
It needs repeating, however, that this group should by no means be seen as a uniform collective, no more than the majority group of "all-Swedish" test takers (born in Sweden with Swedish as their home language).For one thing, this is clearly suggested by the contrast between the good overall results of this group as a whole and the simultaneous over-representation of these students, noted above, within "the small group of distinct low-achievers" (Erickson & Lagergren 2004: 158).Further, on account of their widely differing backgrounds, linguistically and socioculturally, the "foreign" group stands out as much less homogeneous than the "all-Swedish" majority group.For this reason alone, the results of this group are worth studying in more detail, paying due attention to the various subgroups contained in it.
Within the "foreign" group, about ten per cent of the Swedish test takers (cf.above), some 35 different home languages are represented; the number of countries of birth is around 50 (including English-speaking countries), the most frequently stated of which are Somalia, Iraq and Bosnia (Skolverket /Erickson 2004: 63). 6In more specific numerical terms, among the 1431 students making up the total number of test takers in Sweden, Swedish is the stated home language of 1218 students, regardless of country of birth (including, among others, students adopted from abroad at an early age, altogether 49).Among these 1218, in turn, 1166 are "all-Swedish" (born in Sweden, home language: Swedish).The total number of students with another home language than Swedish is 152.This figure includes nine students with English as their home language and one with sign language.Since these ten test takers were not considered representative of the group targeted in this study, they were excluded, leaving a "non-Swedish" group of 142 students, born in or outside Sweden, with another language than Swedish as home language. 7s will have appeared, the "non-Swedish" group of 142 students represents a number of different home languages as well as countries of birth.While the collective results of this group are clearly of interest in themselves, e.g. in comparison with those of the "all-Swedish" group, it should also be of considerable relevance, in view of the aims of this study, to compare results between different home-language subgroups within the larger group, as well as between Swedish-born and non-Swedish-born students within each subgroup.Therefore, seven reasonably large home-language groups, ranging from 27 to a "pragmatic minimum" of eight students each, were set up.They are, in order of decreasing numbers: Arabic, Spanish, Somali, Farsi/Dari, Kurdish, Turkish, and Bosnian/ Croatian/Serbian (Table 1 Altogether, the 88 students in the seven home-language groups make up 62% of the group of 142 "non-Swedish" students in focus here.As shown in Table 1, the students born in Sweden are heavily outnumbered by those born elsewhere: 26 (29%) versus 62 (71%).As can also be gathered from Table 1, the home-language groups differ substantially not only with regard to size (cf.e.g.Arabic versus Turkish) and proportion of Swedish-born versus non-Swedish-born students (cf.e.g.Farsi/Dari versus Kurdish), but also when it comes to the relative "distance", whether in terms of "genetic" or typological relatedness, between a particular home language and English (or Swedish).Further, along another dimension, there are differences as to country of birth, not only between home-language groups, such as Spanish and Somali, both also within them.A case in point is Arabic, corresponding to a number of stated countries of birth (e.g.Algeria, Egypt, Iraq, Kuwait, Syria).This, however, will not concern us here, despite possibly relevant contrasts between colloquial dialects of Arabic as opposed to the "pan-Arabic" or "standard" variety (cf.Kaye 1987:666f.;Smith 2001:195: WALS 2005).
Admittedly, even though a minimum of eight students per home-language group was set as a lower limit, the number of students representing each home-language group is not very large (apart from the Arabic group).Consequently, comparative results between groups will have to be treated with caution.A further source of uncertainty, inherent in most test material, is that a number of students have failed to answer all test items (cf.note 2).A different kind of caveat derives from the inconvenient fact that, as already mentioned, the basic material provides no data on the age of arrival in Sweden for the non-Swedish born students, clearly a relevant factor with regard to command of Swedish, amount of English teaching received and exposure to English.This means that the material will allow no discussion or conclusions concerning the respective roles of "learning" versus "acquisition" (in the sense of, e.g., Krashen 1981) for different groups of students, i.e. beyond the assumption that many students born outside Sweden and with other home languages than Swedish will, generally speaking, not only have had less English teaching than their Swedish-born peers, but also less exposure to English at large.Nor will it be possible, in discussions of possible cross-linguistic influence, to make a principled distinction between what Jarvis & Pavlenko (2008:175) refer to as "learning-related" and "performance-related" effects on transfer.(For reviews and discussion of related issues, see DeKeyser 2003, Odlin 2003:476 (implicit versus explicit learning) and Hulstijn 2003 (incidental versus intentional learning).) In the next section, accounting for the main quantitative results of the investigation, the groups specified in this section -not only the seven homelanguage groups -will provide the basis for comparison.My main focus will be on the "non-Swedish" group, whether as a whole or as represented by the homelanguage groups, especially in relation to the "all-Swedish" group.Throughout, attention will be paid to the distinction, within the "non-Swedish" group, between those born in Sweden and those born elsewhere.9

Overall test results
Before we proceed to the core of the present study, i.e. the outcome concerning (written) linguistic proficiency/correctness and written production, something should be said about the results of different groups on the Assessment of English test as a whole.We have already seen that the overall result of students with a "foreign" background (born outside Sweden and/or with another home language than Swedish) is lower than the total Swedish result, but only to a relatively minor degree.However, to get a more nuanced picture, let us first consider all the tasks on all the skills tested (including receptive skills like listening and reading), including the exercises and items that were later dropped (see section 2).The following groups of test takers will be taken into account: the whole group of test takers, regardless of home language and country of birth (n=1431); the group of test takers with Swedish as their home language, regardless of country of birth (n=1218); the "all-Swedish" group: Swedish as home language, born in Sweden (n=1166); the "non-Swedish" group: home language other than Swedish, regardless of country of birth (n=142); and -for comparison -the "English" group: English as home language, regardless of country of birth (n=9).Table 2 provides an overview of the groups involved, the maximum number of aggregate points on the ten different exercises being 79:

"mean"= average number of points out of 79 awarded for correct answers on the ten exercises;"%"= percentage of correct answers in relation to maximum number of points (79)
Table 2 shows, first of all, that the overall result of the "English" group is well ahead of the others, with 86% correct answers, to be compared with, e.g., 71% for the "all-Swedish" group.Any other outcome would, of course, have cast grave doubt on the validity of the Assessment of English test for measuring general English proficiency.Not unexpectedly, the differences between the whole group, the Swedish-as-home-language group and the "all-Swedish" group are negligible (or non-existent).The main reason why the result of the whole group is (slightly) lower than for the other two is that this group, but not the others, includes the "non-Swedish" group of 142 students.
Apart from the "English" group, the "non-Swedish" group is the one that stands out in Table 2. Its overall result (51.6 points out of 79) is considerably lower than those of the other groups, 65% versus 70-71% correct answers.This is a substantial difference, well worth looking further into. 10For one thing, differences between home-language groups should be investigated, with regard to both overall result and the five exercises dealing with written proficiency in a narrow sense.
Table 3 accounts for the overall results of the seven "non-Swedish" homelanguage groups established in section 3; the "all-Swedish" group serves as a basis for comparison.As is readily seen from Table 3, there are obvious differences between the seven home-language groups' overall results on the Assessment of English test.Three different clusters may be discerned.The results of the Spanish, Farsi/Dari and Bosnian/Croatian/Serbian groups, the top three, are more or less identical to that of the "all-Swedish" group -all of them at roughly 70% correct answers.There is an intermediate group, consisting of Arabic and Turkish (62% and 61%, respectively), well below the "all-Swedish" group and the top three homelanguage groups.The lowest results, in turn well below those of Arabic and Turkish, are to be found for the Somali and Kurdish groups, with just over half of the answers correct.
The results so far thus indicate clear differences in overall results between the seven home-language groups, where some are level with the "all-Swedish" students, others lagging behind.This shows, among other things, that labels like "foreign" or "immigrant background" tend to obscure the heterogeneity of the underlying reality: home-language groups may, for whatever reason (linguistic, sociocultural, etc.), exhibit great variation among them.However, differences within home-language groups should also be taken into account.In particular, as already noted, age of arrival in Sweden would seem to be an important factor.Since such background information about the students participating in Assessment 10 In this connection, the very modest difference in overall result pointed to above between the whole group and the "immigrant background" group (cf.section 3) should be related to the fact that the latter group includes not only the "non-Swedish" group of 142 students, but also the group of 51 students who, while stating Swedish as their home language, were not born in Sweden (adopted at an early age, born to Swedish parents working abroad, etc.); it also includes the nine students with English as their home language.
of English was not provided, we shall have to confine ourselves to what is available, viz.country of birth, already put to use in establishing the "all-Swedish" versus the "non-Swedish" groups (section 3).Let us now consider differences within the home-language groups along the dimension of "Swedishborn" versus "non-Swedish-born" (cf.Table 1), as accounted for in Table 4: In view of the fact that in some home-language groups (Somali, Kurdish and Bosnian/Croatian/Serbian) there is no or just one Swedish-born test taker, the overview provided in Table 4 does not allow any far-reaching conclusions.Even so, disregarding the groups just mentioned, the general tendency seems clear, viz.that Swedish-born students outperform those born elsewhere.For what it may be worth, given its limited number, the Turkish group displays the most drastic difference, the four Swedish-born students achieving 80% correct answers, almost twice the percentage of the four born outside Sweden (42%).The corresponding difference within the Arabic group is also substantial (74% versus 58%), more so than within the Spanish and Farsi/Dari groups, where the difference is roughly the same (8% and 7%, respectively).It is of particular interest to note that the average results of the 24 Swedish-born students within the Arabic, Spanish, Farsi/Dari and Turkish home-language groups are all higher than the result of the "all-Swedish" group (71%; Table 3).The opposite holds for the 35 non-Swedish-born students within the same home-language groups.If we add the non-Swedish-born students in the remaining home-language groups, this tendency is considerably strengthened, with the notable exception of the Bosnian/Croatian/Serbian group, where the result of the seven non-Swedish students (70%) almost equals that of the "all-Swedish" group.
Thus, with regard to the overall results of the "non-Swedish" group on the whole Assessment of English test, the heterogeneity of this group is striking.There are great differences both between and within the various home-language groups, where the dimension of "Swedish-born" versus "non-Swedish-born" looms large.Despite the restricted number of test takers in the seven homelanguage groups, these differences are of considerable interest, not least in relation to the result of the "all-Swedish" group.Keeping the overall picture in mind, we now proceed to the results of the home-language groups on the tasks devoted to written proficiency in a narrow -or perhaps impoverished -sense (cf.section 2).

Results on written-proficiency tasks
As specified in section 2, the five exercises (4, 5, 7, 9, 10) investigated here, aimed at "linguistic competencies" and "written production" (henceforth: "written exercises"), comprise 41 points, i.e. just over half of the total number of points (79) in the full Assessment of English study.The results of some specific groups on these written tasks are accounted for in Table 5: In comparison with the corresponding results for the whole test (Table 2), it may be noted that the percentages of correct answers (out of the maximum 41 points) are considerably lower for the five written exercises, i.e. these tasks come out as more difficult than the test as a whole.Further, here as well as for the overall results, we find the "English" home-language group well ahead of the others; in fact, the difference in relation to the second best group, the "all-Swedish" one, is somewhat more pronounced (17%) than for the whole test (15%; Table 2).This may be seen as a further indication that the level of relative difficulty is higher for the five written exercises than for the whole test.
Table 5 also shows that, as in the overall results, there is hardly any difference between the results of the whole group, the Swedish home-language and the "all-Swedish" groups as far as the writing tasks are concerned (62-63% correct answers).Considerably lower results are shown by the group of "non-Swedish" students (57% correct answers), the distance to the "all-Swedish" group (6 percentage points) being the same as that in the overall results (cf.Table 2).
Having established a clear difference between the "all-Swedish" group and the "non-Swedish" group in its entirety, let us now consider the separate results of the seven home-language groups on the five written tasks, as accounted for in Table 6: Again, as was the case for the overall results (Table 3), we find the seven homelanguage groups presenting a heterogeneous picture, results on the written tasks ranging from 41% to 65% correct answers (corresponding range for overall test results: 51-72%).The results of the top group of Spanish, Farsi/Dari and Bosnian/Croatian/Serbian (62-65%) are close to or (for Bosnian/Croatian/Serbian) even above that of the "all-Swedish" group.An intermediate group consists of Turkish and Arabic (50% and 54%, respectively).The lowest results on the written exercises are those of the Somali and Kurdish groups (41% and 45%, respectively).These three groupings are identical to those for the overall results.
On the whole, then, the results of the seven home-language groups on the five written exercises correspond quite well to their overall test results, both compared to the "all-Swedish" group and with regard to differences between them.However, as shown for the overall results, there is reason to believe that there may also be differences within the home-language groups, as was found to be the case for the overall results.Table 7 is focused on the distinction between those students among the home-language groups who were born in Sweden and those who were not: Generally speaking, as can be seen from Table 7, the results of the Swedish-born students (in groups with at least four such students) are considerably higher than those of their non-Swedish-born peers in the same home-language groups.This finding is well in agreement with the overall test results (Table 4).The main difference is that, due to the greater difficulty of the written exercises in comparison with the whole test, percentages of correct answers are lower on the written tasks.Thus, the results of the Swedish-born students in the Farsi/Dari, Spanish, Arabic, and Turkish home-language groups are in the range 65-72% correct answers on the written tasks, to be compared to 74-80% on the whole Assessment of English test.
As in the overall outcome of the test, it is worth noting that the average results of the 24 Swedish-born students within these home-language groups are all higher than the result of the "all-Swedish" group (63%; cf.Table 6), as opposed to the 35 non-Swedish-born students within the same home-language groups.The difference between Swedish-born and non-Swedish-born students is especially striking within the Arabic and Turkish groups.If we add the non-Swedish-born students in the remaining home-language groups, the general tendency is further strengthened -with the notable exception of the Bosnian/Croatian/Serbian group.The average result of the seven non-Swedish-born test takers in this group (65% correct answers) actually surpasses that of the "all-Swedish" group.On the writ-ten tasks, then, these non-Swedish-born students perform on a par with the Swedish-born students in the Farsi/Dari, Spanish and Arabic groups.
So far we have considered aggregate results on the five written tasks.Table 8 gives the results of the seven home-language groups on each of the five written exercises (for a brief description of each such exercise, see section 2).Owing to the differences in maximum number of points between different exercises (range: 4-16), the distribution of results differs a great deal between them, also in terms of average number of points and percentages of correct answers.For instance, on ex. 4 (4 points), the Swedish-born students in the Arabic home-language group have 95% correct answers, the non-Swedish-borns 83%; the corresponding figures for ex. 10 (16 points) are 58% versus 38%.The same applies, to an even higher degree, to the Turkish home-language group: compare, e.g., 70% (Swedish-born) versus 58% (non-Swedish-born) on ex. 5 (4 points), to be contrasted with 79% versus 25% on ex.7 (12 points), not to mention ex.10.
The possibility that such differences could be partly due to test-format differences between tasks (multiple choice versus gap-filling) cannot, of course, be a priori excluded.On the other hand, there are no similar differences to be observed within the Spanish group, where the range of variation between different exercises is, on the whole, much more limited.Table 8, supplying much of the nitty-gritty of the findings accounted for earlier, provides ample material for a more detailed investigation of the differences in results on each of the written exercises, both between different home-language groups and between the "all-Swedish" group and specific home-language groups.In particular, it would have been of interest to study, as far as the limited material would allow, the results on different exercises concerning the relationship between their "content" (grammar versus vocabulary, the articles versus irregular verbs, etc.) and specific home-language groups.Such an extensive undertaking, however, is beyond the scope of this article.My aim in the next section will be more restricted, based on just one of the written exercises, viz.ex. 5, dealing with the English articles.The general question focused on is this: to what extent is it possible to find a connection between a specific home language and results on individual tasks?In other words, is it possible to find evidence of cross-linguistic influence on the basis of a close study of the results of different home-language groups on ex.5? 5 Cross-linguistic influence?5.1 Background Since the mid-20th century, ideas about cross-linguistic influence -or transferhave travelled a long and somewhat winding road. 11As noted by, e.g., Ringbom (1987:50), teachers have probably always been aware that similarities and differences ("contrasts") between languages -especially, but not exclusively, between a first language (L1) and a second or foreign language (L2) -play some kind of role for students learning a new language.In more general terms, Jarvis & Pavlenko (2008:1) state that cross-linguistic influence "has been of interest to laypeople and scholars alike since antiquity and most likely ever since language evolved."However, in Kelly's (1969:54) words, "only after the renaissance did teachers exploit the differences between languages in teaching them."Thus, in a basic sense, "contrastive linguistics was born of classroom experience" (Bolinger 1971:vii).In this connection, the strongly stated conviction of Robert Lado (1957:2), the credo of early contrastive analysis, is well known, viz.that for the language student, "[t]hose elements that are similar to his native language will be simple for him, and those elements that are different will be difficult" (cf.Selinker 1992:9ff.).The relevance of such ideas was called into question, if not downright denied, in the 1970s and much of the 1980s, largely but not exclusively based on evidence from the study of what appeared to be universal developmental sequences in language acquisition, regardless of the L2 and L1 involved; "L2 acquisition = L1 acquisition" (Dulay & Burt 1974:107). 12The contrastive perspective was defended by, among others, Fisiak (1981:7), arguing that "[t]he value of contrastive studies lies in its [sic] ability to indicate potential areas of interference and errors.Not all errors are the result of interference." 11 For discussion of the terms "cross-linguistic influence", "transfer" and "interference" -the latter two with a historical (and somewhat stigmatized) connection to behaviourist learning theory -see Ringbom (1987:1, 46), Odlin (1989:25ff.),Gass & Selinker (2001:118f.),Jarvis & Pavlenko (2008:3).Somewhat unusually today, "interference" is used in the subtitle of Swan & Smith (2001). 12Cf.Selinker (1992:172): "Lado's position can be viewed as one extreme language transfer position, the view that second language learners rely almost exclusively on their [native language] in the process of learning a particular [target language].With attacks on this position gaining currency in the early 1970s ..., the other extreme of the transfer hypothesis came to the fore: the view that language transfer is unimportant ..." Selinker (p. 3) refers to the situation as an instance of the "infamous 'baby and bathwater syndrome'".On the "historical skepticism about transfer", see also Jarvis & Pavlenko (2008:8ff.).
As the 1980s were drawing to a close, scepticism towards transfer and contrastive considerations as a relevant factor in language learning -especially with regard to morphology and syntax (Odlin 2003:439ff.)-had gradually given way to a re-evaluation.This included a shift of attention from negative to positive transfer, to the role of similarity rather than difference as a source of some types of transfer (Odlin 1989:113f.),but where the underlying assumption was: "The learner of a closely related L2 knows a great deal about the language even before he has started learning it" (Ringbom 1987:53;cf. Ringbom 2007).
According to Odlin (1989:152), "[t]ransfer occurs in ALL linguistic subsystems", "both in informal and formal [learning] contexts."Odlin also criticizes "the assumption that theories of transfer are inextricably linked to [behaviourist] theories of habit formation" (p.23).Further, "there are reasons to believe that cross-linguistic influences work in tandem with the psychological factors governing developmental sequences" (p.23), an idea in line with contemporary thinking on the subject of cross-linguistic influence.In a similar vein, Gass & Selinker (1992:7) point to current efforts "to reconcile a language transfer perspective and a cognitive perspective, in general ... and a language transfer perspective and a developmental perspective, in particular".Summarizing research on transfer and related matters, they emphasize that "[t]here is now overwhelming evidence that language transfer is indeed a real and central phenomenon that must be considered in any full account of the second language acquisition process" (p. 7). 13 This is not the place to discuss the complex issues concerning cross-linguistic influence in L2 learning; "there does not yet exist any comprehensive theory of language transfer" (Odlin 2003:478;cf.Gass & Selinker 2001:117ff., 185ff.;Jarvis & Pavlenko 2008).My point of departure here is simply that, whatever its exact nature and whatever constraints it is subject to, such influence exists: "L1 acts as a major factor in L2 acquisition" (Ellis 1994:343), at different levels (formal, semantic and conceptual; cf.Odlin 2003:464ff.);Jarvis & Pavlenko 2008:112ff.), involving individual items (e.g.specific endings or prepositions) or more abstract, structural properties of a language (e.g.article or tense systems).In the words of Ringbom (1987:2), the basic question no longer concerns the existence of transfer phenomena, but "in what circumstances L2 learners transfer what; how much is transferred and why."At the same time, it is important to bear in mind that far from all L2 learning problems are due to cross-linguistic influence (cf.e.g.Köhlmyr 2003).
One problem in this context is that transfer is not always easy to prove -or indeed to falsify -in individual cases. 14 Gass & Selinker (2001:183) emphasize 13 Cf.Ellis (1994:341): "No theory of L2 acquisition is complete without an account of L1 transfer."A clear indication of this is the appearance of Odlin's (2003) lengthy chapter in The Handbook of Second Language Acquisition (Doughty & Long 2003).Twenty years earlier, such a chapter in such a book would have been unlikely. 14 Cf. Selinker (1992:208), referring to what he sees as a "key question": "How does the analyst unambiguously show that language transfer has occurred?"For further in-depth discussion of different types of cross-linguistic evidence, see Jarvis & Pavlenko (2008:27ff.).
that "[u]nlike L1 grammars, no two individuals have the same L2 grammar and hence there is no way of predicting what will happen to a grammar when new information is added, causing changes in the existing system."Therefore, due to interlanguage variability, consistency in, for example, test results is not to be expected even among students sharing the same L1.At the same time, among language teachers in general, as opposed to linguists in the 1970s, the pedagogical relevance of a (partly) contrastive perspective seems never to have been in serious doubt.Further, contrastively orientated school grammars, adhering to a long pedagogical tradition, can be said to epitomize generations of teacher experience as regards language differences known to cause problems to learners with a specific L1.
For example, as teachers of English in Sweden will know, Swedish learners of English are apt to produce sentences like *I don't know how many of the students that speak Arabic, where the insertion of that as a sort of subject marker in the interrogative clause derives from the corresponding Swedish structure (... hur många av studenterna som kan ...) -a type of error far less likely to be committed by, for instance, a German or French learner of English.For Swedish learners, it makes eminent sense to stress this structural difference between English and Swedish.In today's multilingual Swedish society, however, contrastively (and empirically) based "warnings" need a broader, more multicontrastive basis than just one specific L1.This is the kind of context in which the present study should be seen. 15y and large, the "chronic" difficulties in establishing transfer effects are also inherent in the data provided by the seven home-language groups under special scrutiny here.Two things, in particular, speak against any hard and fast conclusions as regards cross-linguistic influence: the limited number of non-Swedish-born students in each home language group and the specific nature, or test formats, of the five exercises investigated.As already pointed out, tasks like those used in Assessment of English, involving multiple choice or gap filling, are not ideal for tracking down cross-linguistic influence, whether of a direct or more indirect kind; some answers, and errors, may in fact be partly due to test-format effects.By contrast, free written (or, of course, oral) production, by a considerably larger number of "non-Swedish" students, would have made possible a more solidly based and comprehensive study of not only systematic grammatical features (like word order, tenses, presence/absence of subject pronoun or copula, etc.), but also of more subtle or "indirect" transfer effects, like avoidance or overrepresentation of certain structures (cf.Gass & Selinker 2001:119ff.).Thus, given the limitations of the material, the discussion of cross-linguistic influence in the next subsection should be seen as in large measure tentative.
In this connection, another pertinent circumstance, closely interwoven with the composition of the "non-Swedish" home-language groups, is that there may well be transfer effects not only from the home languages in question, but also from L2 Swedish, i.e. so-called interlanguage transfer, also referred to as "lateral" transfer (see Odlin 2003:477;Jarvis & Pavlenko 2008:12, 21f.). 16As pointed out by Gass & Selinker (2001:134), interlanguage transfer gives rise to questions like these: "Does the gradually acquired knowledge of a language beyond the second make a difference in the types of transfer seen?How is knowledge of a prior interlanguage used (or not used) in the addition of a third, fourth, or fifth language?"Such questions, to be touched on in the next subsection, are highly relevant to the data studied here, further complicating the overall picture.However, exploring them in a systematic way is beyond the scope of the present study.Suffice it to say, for now, that it is reasonable to assume that the relatively short language distance between Swedish and English -as opposed to, say, English and Arabic or Turkish -may imply positive transfer effects for "non-Swedish" students with a fairly good command of Swedish, especially when born in Sweden.

The articles
In the following discussion of possible cross-linguistic influence in the Assessment of English data, I shall focus, in some depth, by means of an informal kind of error analysis, on a central aspect of English grammar, viz. the use of the indefinite and definite articles, as tested in ex. 5.This means that items concerning, for example, indefinite quantifiers, like any and every (ex.4), the forms of ir-regular verbs (ex.7) or the do-construction (ex.9) will not be dealt with; nor will the choice of prepositions or gaps of a purely lexical nature (ex.10).
Ex. 5 consists of the following four multiple-choice items (correct answers underlined; cf.Bonnet 2004:170):17 5:1.They are fond of ___ science fiction films.(the, a, "zero", an) 5:2.I've bought you ___ interesting book.(a, "zero", an, some) 5:3.Did you like ___ book I gave you? (an, the, "zero", a) 5:4.In my new school there are ___ very young teachers.("zero", an, the, a) Let us start our discussion with items 5:1 and 5:4.In both of them, the correct answer is the "zero" article before the plural nouns, signalling indefinite form.
In the Arabic home-language group (n=27), a fair number of students have chosen the indefinite singular article (a or an) instead of "zero" as the answer in both items.On 5:1, out of the 21 non-Swedish-born students, six (29%) have opted for a/an, as opposed to none of the six Swedish-born ones.For item 5:4, a similar picture emerges: six non-Swedish-born students and two of the Swedishborn ones have chosen a/an.
These results seem to indicate that especially non-Swedish-born students in the Arabic group find it difficult to come to grips with the English article system, in particular the use or non-use of the indefinite and "zero" articles.For their Swedish-born peers, the situation looks somewhat different.It is tempting to interpret these results as at least partly due to cross-linguistic influence, i.e. from Arabic, the home language of these students.The general situation in relation to English is described as follows by Smith (2001:205): "There is no indefinite article in Arabic, and the definite article has a range of use different from English."For an Arabic-speaking learner of English, this means learning not only the exponents (a/an, "zero") of a specific grammatical category (indefiniteness), but also -more importantly -the functional "point" of this category.There is no "free ride" from Arabic to benefit from, unlike, e.g., the situation for learners of English with Swedish as their L1, where the article system is, in broad outline, more or less identical to that of Swedish (cf.Ljung & Ohlander 1993:8, 37).Ringbom (1987:136), re-ferring to the learning task at hand as an "organizational" problem, rather than one involving merely the choice between different items of the same category (e.g. a versus an), points out that "[w]hen there is little system similarity between L1 and L2 the learner is faced with a number of organizational problems."In view of this, it is logical that "[t]he indefinite article causes the most obvious problems ..." for Arabic-speaking learners and, further, that "[w]hen the English indefinite article has been presented, it tends to be used wherever the definite article is not used: *These are a books" (Smith 2001:205).The close resemblance between this kind of error and those under discussion here is readily seen.
It seems likely, therefore, that the choice of a/an instead of "zero" to fill the gap in front of plural nouns may have been influenced by the contrast in article systems between English and Arabic, resulting in a sort of "functional/structural confusion" for Arabic-speaking learners when faced with a new type of articleand thus an unfamiliar grammatical distinction -in the target language.The fact that this kind of transfer effect is considerably less pronounced among the Swedish-born students in the Arabic home-language group may be related to the obvious likelihood that these students, due to their being born in Sweden, will as a rule have had more time, and abundant opportunities, to learn Swedish as a second language as well as studying English in the Swedish school and being exposed to English as part of their everyday life, like most learners of English (and other people) in Sweden.Not least, their presumably greater command of Swedish, compared to those not born in Sweden, will have enabled them to benefit more fully from the structural similarities between English and Swedish.This may well be an instance of interlanguage transfer in the sense noted above, involving the interplay of not two but three languages.
The uncertainty among, in particular, the non-Swedish-born students in the Arabic group concerning the use of the indefinite (including "zero") article also manifests itself, albeit less drastically, in ex.5:2:.I've bought you ___ interesting book (a, "zero", an, some).Eleven (52%) out of the 21 non-Swedish-born students have chosen the correct an (five have chosen a, one "zero" and three some; one: no answer); the corresponding figure for the six Swedish-born students is five (83%) (one has chosen a).
Interestingly, however, ex.5:3 -Did you like ___ book I gave you? (an, the, "zero", a) -turned out to be almost completely unproblematic: only one student (non-Swedish-born) in the entire Arabic home-language group (n=27) has failed to choose the definite article.It should be noted that this task differs from the other three in ex. 5 in that its focus is on the definite article.This is clearly relevant in view of the circumstance that even though, as quoted above, "the definite article has a range of use different from English", there is such a thing as a definite (in contrast to indefinite) article in Arabic: "It is used, as in English, to refer back to indefinite nouns previously mentioned" (Smith 2001:205).Note that these requirements for the basic use of the definite article in English and Arabic are met by the sequence Ex.5:2-5:3, i.e.I've bought you an interesting book.Did you like the book I gave you?.This, then, may explain why ex.5:3 caused virtually no problems.A contributing factor may have been the functional similarity between English and Swedish concerning the basic use of the definite article in the two languages (cf.Ljung & Ohlander 1993:37, 44), thus quite possibly a case of positive interlanguage transfer from Swedish to English for the Arabic home-language group.In short, both the L1 (Arabic) and the L2 (Swedish) may here be seen as working in the same direction to facilitate the learning of the English definite article.
The above discussion of the items in ex. 5, centred on the possibility of crosslinguistic influence, may provide a clue to the difference in results on this task -3.0 (75%) versus 2.5 (63%) -between the Swedish-born and the non-Swedishborn students in the Arabic group (Table 8).It should be noted that the result of the Swedish-born subgroup is identical to that of the "all-Swedish" group of test takers, supporting the idea that the greater knowledge of Swedish on the part of the Swedish-born Arabic-speaking students gives them an edge in learning English in comparison with their non-Swedish-born peers.Extrapolating the discussion of cross-linguistic influence to other areas tested in the Assessment of English survey may help to explain the difference in overall results on the five written exercises between the Swedish-born and the non-Swedish-born Arabicspeaking students in comparison with the "all-Swedish" group, accounted for in Table 8 (67%, 51% and 63%, respectively).Again, it seems reasonable to assume that the overall difference between the two Arabic-speaking subgroups -as well as the similarity in results between the Swedish-born subgroup and the "all-Swedish" group -could be related to knowledge of Swedish.The fact that the Swedish-born Arabic-speaking students actually perform better than the "all-Swedish" group will be briefly touched on in the concluding section.
The picture of cross-linguistic influence emerging in above discussion of the Arabic home-language group may, at first sight, seem fairly convincing.However, it may equally well be thought of as too rosy.Lest we forget: the tasks have certain limitations, numbers are small and, further, the Swedish-born versus non-Swedish-born distinction also entails other than purely linguistic differences (social and educational, such as number of years in the Swedish school system, studying English, etc.), which may be of great relevance in this connection, possibly interacting in complex ways with cross-linguistic influence.Further, the overall difference in proficiency level, as shown in the whole Assessment of English test, between the Swedish-born and the non-Swedish-born test takers (cf.Tables 4, 8) may in itself imply a lower result on ex.5:1-4 (cf.Jarvis & Pavlenko 2008:201ff.).On the other hand, as already suggested, it may -with equal justification -be argued that cross-linguistic influence of the kind discussed so far in fact provides a partial explanation for the proficiency difference between the two Arabic-speaking subgroups.As emphasized above, proving transfer effects is not always easy, and so a good deal of caution is called for.With this in mind, let us briefly consider some other home-language groups in relation to ex. 5. First, however, a few words on language distance are in order.
There can be little doubt that, from a typological point of view, the general distance between English and Arabic is far greater than that between English and Swedish, not only when it comes to articles.In his concluding discussion of transfer, Odlin (1989:153) states that "[l]anguage distance is a factor that affects transfer" and also that "[t]ypological factors can affect the likelihood of transfer" (cf.Odlin 2003:471, Jarvis & Pavlenko 2008:176ff.). 18According to Odlin (1989: 39), one experience-based indication of the assumed effect of relative language distance on language learning consists in the maximum number of weeks, in intensive language courses at the Foreign Service Institute at the U.S. State Department, required for an English-speaking student to develop "linguistic skills to a high level of proficiency that is comparable in each of the languages".For example, to learn Swedish would require 24 weeks versus 44 weeks for Arabic; for Turkish, too, the number of weeks is 44, as opposed to only 20 for Spanish.This means that, for the average speaker of English, Swedish as well as Spanish is supposed to be considerably easier to learn than both Arabic and Turkish, largely due to relative language distance, i.e. the distance measured in terms of similarity and difference between, say, Arabic and English, or Swedish and English.The same would apply in the opposite direction, for speakers of Arabic or Swedish learning English.
Like Arabic, Turkish differs in important respects from English as regards articles.In particular, there is no definite article in Turkish (WALS 2005), although a specific suffix (-i) is added to nouns with definite reference serving as direct object (Lewis 1953:27, Thompson 2001:223).On the other hand, Turkish does have an indefinite (singular) article, "the same as the word for 'one'" (Lewis 1953:26;cf. Thompson 2001:223).Thus, the Turkish situation as regards articles is almost a mirror image of the Arabic one (definite article, no indefinite article; cf.above) -but just as different from the English system.
The results of the Turkish home-language group (n=8) on ex. 5 are slightly lower than for the Arabic group (Table 8): Swedish-born: 2.8 (70%); non-Swedish-born: 2.3 (58%).The general tendency, it should be noted, is the same as for the Arabic group, the four Swedish-born students performing clearly better than the four non-Swedish ones.Again, it is a reasonable conjecture that a better knowledge of Swedish among the former subgroup may be part of the explanation for this difference.A closer look at some individual items in Ex. 5 reveals that three out of four in the non-Swedish subgroup have chosen a/an as the correct answer in 5:1 (They are fond of ___ science fiction films); none of the four in the Swedish-born subgroup has.On the other hand, in ex.5:4 (In my new school there are ___ very young teachers), the non-Swedish-born students do slightly better (75% correct, i.e. "zero", answers) than the Swedish-born ones (50%); in each subgroup, one test taker has opted for a.In ex: 5:2 (I've bought you ___ interesting book) the situation is the reverse: the Swedish-born students have all chosen a/an, whereas one of the non-Swedish students has chosen "zero", as is also the case in item 5:3 (Did you like ___ book I gave you).
The results of the Turkish home-language group suggest that the English article system, as tested in ex. 5, seems to give rise to considerable learning difficulties, at least in part due to differences between Turkish and English.However, in view of the clear-cut structural differences between Arabic and Turkish as regards articles, it is interesting to observe that these two home-language groups appear to have partly similar problems with English article use, such as choosing the indefinite singular article (a/n) in front of plural nouns.In both groups, this applies especially to the non-Swedish-born students.This may be taken to indicate that, at least when it comes to learning the English article system, it is probably vain to look for point-by-point transfer between L1 and L2, to the effect that the existence or non-existence of either the indefinite or the definite article in an L1 will automatically facilitate or complicate the learning of the corresponding article in English.Instead, cross-linguistic influence when it comes to articles may be thought of as operating more indirectly, at a more abstract system level, where perceived similarities and differences in the overall structure of article systems may be what really counts, creating a sense of relative familiarity or confusion in the minds of learners. 19In this regard, learners of English having Swedish as their L1 should find it easier to grasp, or recognize, the overall function and structure of the English article system than, for example, learners having Arabic or Turkish as their L1.At a basic structural level, as pointed out earlier, the distance between, for instance, the Swedish and the English article systems is small; the system at large will appear more or less transparent.If, on the other hand, essential pieces are found "missing" from one's L1 in relation to the target language, then, obviously, the perceived distance may seem almost overwhelming.The article system will appear more or less opaque, resulting in functional as well as structural confusion, particularly at early stages of learning.This may well be a matter not only of objective linguistic differences but also, perhaps even more so, of a highly subjective interpretation of the cross-linguistic differences and thus of the learning task at hand. 20panish is, of course, typologically much closer to English than are languages like Arabic and Turkish.Further, like Swedish, Spanish has an article system that is structurally fairly similar -although with certain differences (Coe 2001:104; cf. also below) -to that of English, with both indefinite and definite articles.Within the Spanish home-language group (n=13), too, there is a clear difference between the two subgroups as regards ex. 5 (Table 8).The score of the seven Swedishborn students is 3.3.(83% correct answers) -the highest (but cf.below) of all the home-language groups, also higher than the "all-Swedish" group at 3.0 (75%).The corresponding result of the six non-Swedish-born students is 2.7 (68%).It may be noted that all the students in the Spanish group have chosen a/an in item 5:2 (I've bought you ___ interesting book) and also that they correctly opted for the in the follow-up item 5:3 (Did you like ___ book I gave you).This may, perhaps, be partly attributed to positive transfer from Spanish.As to item 5:1 (They are fond of ___ science fiction films), one non-Swedish-born student has incorrectly chosen the indefinite article a; in item 5:4 (In my new school there are ___ very young teachers), half of the six non-Swedish students have opted for a/an, as opposed to only one of the seven Swedish-born students.The erroneous choice of a/an before plural nouns, it will be remembered, was also found to be a feature of the Arabic and Turkish home-language groups.For the Spanish group, however, there may be a more specific contrastive reason; what may look like the same kind of error may have multiple causes. 21As noted by Coe (2001:104), "[t]he indefinite article has a plural form", which "can cause beginners to make mistakes", i.e. in English.This may well have contributed to the choice of a/an in front of plural nouns by students in the Spanish home-language group.The fact that, in the limited material investigated here, this choice is four times as frequent among the non-Swedish-born than the Swedish-born students may well be a consequence of (positive) interlanguage transfer: their presumably greater knowledge of Swedish may have helped the Swedish-born -in contrast to the non-Swedish-born -students in the Spanish home-language group to counter L1 influence as regards indefinite plural nouns.Like English, Swedish has no overt plural indefinite article.
Against the background of relative language distance and "the head start that speakers of some languages have in coming to a new language" (Odlin 1989:27), the analyses proposed above concerning the role of cross-linguistic influenceincluding interlanguage transfer -in the results of the Arabic, Turkish and Spanish home-language groups may carry some prima facie plausibility.However, in order to get a more complete picture, including some possible counterevidence, let us take a brief and selective look at the results of the remaining home-language groups.Space does not allow a more detailed qualitative study of each item in ex. 5, as was done for the Arabic, Turkish and Spanish groups, only a few cursory observations with some bearing on our main preoccupation.Also, for the majority of the remaining home-language groups (Somali, Kurdish, Bosnian/Croatian/Serbian), the number of Swedish-born students is so low that pursuing the Swedish-born versus non-Swedish-born dimension would be a fruitless endeavour.
The Farsi/Dari group consists of seven Swedish-born and four non-Swedishborn students.The results of these two groups on ex. 5 are identical: 3.0 (75% correct answers), i.e. the same as for the "all-Swedish group" (also the same as for the Swedish-born Arabic-speaking students; Table 8).In view of the typological distance between Farsi/Dari and English, 22 a lower result, especially for the non-Swedish-born subgroup, might have been expected, as in the case of the Arabic, Turkish and Spanish home-language groups.In any event, the results on ex. 5 in the Farsi/Dari group do not in themselves provide evidence for transfer.Still, there are, in this group as well, altogether four examples of the incorrect choice of the indefinite article before plural nouns (ex.5:1, 5:4).
The seven non-Swedish-born students (there is only one Swedish-born student) in the Bosnian/Croatian/Serbian home-language group may be regarded as even more exceptional, in that their average result on ex. 5 is 3.3 (83% correct answers), equal to that of the seven Swedish-born students in the Spanish group, to be compared with 3.0 (75%) for the "all-Swedish" group.This result is all the more remarkable as these three Slavic languages, like Polish and Russian, lack both indefinite and definite articles (WALS 2005). 23In such conditions, with no L1 help as far as articles are concerned, problems with the English article system would have been natural enough; there are, in fact, two incorrect answers involving the indefinite article in front of plural nouns.It might be argued, of course, that having no articles in one's L1 means coming to the new language with a sort of blank slate, thus eliminating transfer effects due to treacherous partial similarity.In my view, this kind of argument, not unheard of in the 1970s (cf.Ringbom 1987:52ff.), is of dubious overall validity as it builds on the assumption that, in general, it is similarity rather than difference that gives rise to 22 Cf.Wilson & Wilson (2001:188): "The definite and indefinite articles have no equivalent in Farsi.... The system is quite different from English and Farsi speakers have considerable difficulty with all areas of article use." 23 Cf. the following statement on Russian learners by Monk & Burak (2001:155): "Russian has no articles.The mistakes Russian learners make in the use of the English articles are too numerous to list here.There is a general confusion in the understanding of the basic differences between the, a/an and the zero article."Cf. Spiewak & Golebiowska (2001:174) on Polish learners.Incidentally, three non-Swedish-born Russianspeaking students in Assessment of English had no problems with the English articles in ex. 5. learning problems. 24In the case at hand, it means having to learn a new, abstract system from scratch, rather a daunting task, much more so than simply learning, at the item level, that in English the definite article is the, the indefinite a or an.In any event, whatever the reason may be for the result of the Bosnian/Croatian/ Serbian group, it cannot be cited in support of cross-linguistic influence, even though the distance -certainly the perceived distance (cf.note 20) -between these Slavic languages and English is smaller than for Arabic and Turkish.We see, again, that the predictive power of contrastive linguistics is limited (cf.Odlin 1989Odlin :35ff., 2003:441ff.):441ff.),not only at the level of individual learners.In many cases, there are simply too many interacting factors at work in the learning process -apart from cross-linguistic influence -for us to be able to pinpoint, with any degree of confidence, the cause of a certain outcome.Of course, this applies a fortiori to small groups of learners with different home languages and sociocultural backgrounds, as in the present study.
The two lowest results on ex. 5 among the seven home-language groups are those of the twelve Somali-speaking students, all of them non-Swedish-born, and the eight non-Swedish-born Kurdish-speaking students.The former group achieved 1.7 points (43% correct answers), the latter 2.0 (50%).For both these groups, the distance to English may be seen as roughly the same as for languages like Arabic, Turkish and Farsi. 25By and large, this also applies to the specific field of articles in comparison with English.Somali has no indefinite article and no preposed definite article of the English type although there is a definite affix; Kurdish has an indefinite and a definite affix but no English-type articles (WALS 2005, cf. Tingbjörn 2001:219).These characteristics, it would seem, do not provide much of a free ride into the English article system.
In neither home-language group has item 5:3 (Did you like ___ book I gave you?) caused much trouble: only one student in each group has opted for another alternative than the correct the.By contrast, items 5:1 and 5:4, both involving plural nouns, has produced a large proportion of incorrect answers.On item 5:1 (They are fond of ___ science fiction films), half of the Kurdish-speaking students have chosen a/an; the corresponding figure for the Somali group is two.For item 5:4 (In my new school there are ___ very young teachers), the same tendency is even more pronounced: five out of the eight students in the Kurdish group have opted for a (none for an), whereas in the Somali group seven out of the twelve students have chosen a/an.
Thus, it appears to be the indefinite rather than the definite article that constitutes the biggest problem for the non-Swedish-born students in the Kurdishand Somali-speaking groups, especially coming to grips with the "zero" article before plural nouns. 26It is tempting to seek part of the explanation in the fact that the home languages of these students, as well as many others in the non-Swedishborn home-language subgroups, are quite different from English with regard to indefiniteness as a grammatica category.Also, it is likely that, as a group, not born in Sweden, they are unable to benefit from the structural similaritiesincluding the "zero" plural article -between English and Swedish, at least not to the same extent as students born in Sweden.
Some further comments are in order on the finding that the "zero" article as a marker of indefinite plural nouns has turned out to be particularly troublesome for many test takers in the home-language groups.As we have seen, cross-linguistic influence at the system level seems to be a relevant factor, especially for students not born in Sweden, not being able, to the same degree as their Swedish-born peers, to draw on the functional and structural similarities between the English and Swedish article systems.However, there may be an additional factor at work here, complicating the learning task of grasping the article system and its specific exponents.This factor, it may be argued, is the minimal perceptual salience of the "zero" article, which is -zero. 27This means that, in comparison with the other, overtly marked exponents of the article system (a/an, the), the "zero" article is likely to escape notice much more easily in incoming speech (and writing).What is not readily noticed may well take longer to establish itself, in the mind of the learner and thus in production, as an item within a grammatical system.To a Swedish-speaking learner, this is not usually a problem, since roughly the same state of affairs obtains in Swedish.For speakers of many other languages, however, discovering what "is not really there" may not be as self-evident as noticing the appearance of -and differences between -items of phonetic substance, especially at the early stages of learning.Putting it differently, making a mental distinction between nothing and the "zero" article may not be the easiest of tasks.28

Concluding remarks
In the preceding sections, it has been pointed out that the material from the Assessment of English survey of 2002 suffers from certain weaknesses in relation to the aims of this study.For one thing, the test format, involving multiple choice and gap filling tasks, is hardly optimal for investigating written proficiency in English among groups of students with immigrant backgrounds, the focus of the investigation.Also, the number of students, Swedish-born and non-Swedish-born, in the seven home-language groups making up the core of the study is limited, ranging from 27 in the Arabic-speaking group to eight in the Turkish and Bosnian/Croatian/Serbian groups.Therefore, the results arrived at must be treated with caution and conclusions regarded as tentative.Even so, I would argue, the findings of this study are relevant to the further discussion of the learning of English by "immigrant students" in Sweden (as well as other countries). 29 From a basically quantitative perspective, it was found that the "non-Swedish" group of test takers (142 students with other home languages than Swedish, regardless of country of birth), had a markedly lower overall result on the whole Assessment of English test than the "all-Swedish" group (Swedish as home language, born in Sweden), but also that results displayed considerable variation both between and within the seven home-language groups.By and large, the difference in results between Swedish-born and non-Swedish-born test takers in the home-language groups was substantial, the Swedish-born students outperforming those born elsewhere.Roughly the same results and tendencies were found in relation to five tasks dealing with "written proficiency" (in a restricted sense).Here, too, clear differences were found between as well as within the home-language groups.
The "non-Swedish" group thus turned out to be strikingly heterogeneous, perhaps not very surprising in view of the widely differing backgrounds, linguistically and otherwise, of the home-language groups studied.However, as pointed out earlier, the lack of data on the sociocultural background of the test takers has precluded any study of such variables in the material, and so the possible influence of such factors on the general test results cannot be taken into account when interpreting the results.
In any event, the distinction between Swedish-born and non-Swedish-born test takers was found to play a pivotal role in the variation found within the "non-Swedish" group.It should also be noted that, in actual fact, the Swedish-born students in some home-language groups (Arabic, Spanish, Farsi/Dari, Turkish) did better than the "all-Swedish" group on both the whole Assessment of English test and the five written proficiency tasks.This finding is not self-evident, leading on to questions concerning the role of specific home languages -and thus of cross-linguistic influence, from home languages as well as from L2 Swedish -in the process of learning English.
It appears that cross-linguistic influence can be inferred in many of the answers given to the test items targeting the English article system.Although transfer is an area where predictions may prove unreliable and definitive proof is hard to come by, there are different kinds of circumstantial evidence in support of transfer effects in the material analysed (there is also some apparent counterevidence). 29To what extent conscious monitoring (explicit knowledge) and/or intuitive reactions (implicit knowledge) contribute to the test results, e.g. as regards the grammar items, is impossible to tell.Generally speaking, however, the test format for the writing tasks is likely to have elicited more conscious monitoring than, e.g., free written (or, even more, oral) production.Cf. Jarvis & Pavlenko (2003:195), commenting on research suggesting that, if anything, conscious monitoring appears to decrease the occurrence of negative transfer.
One kind of evidence is that home languages whose typological distance to English is great, especially with regard to articles, seem likely to induce types of error -such as opting for the indefinite singular article before plural nouns -that can reasonably be related, often indirectly, to structural properties of the home languages.The transfer effects often seem to be the result of overall confusion at the system level rather than problems of choice at the item level.In short: "Learners whose L1 does not have articles have greater problems learning the English articles than learners whose L1 has articles" (Ringbom 1987:51).
Circumstantial evidence of a different kind is provided by the large difference, in many cases, between the results of the Swedish-born and the non-Swedishborn students in the same home-language groups.This difference is suggestive of interlanguage transfer from L2 Swedish for the Swedish-born as opposed to the non-Swedish-born test takers.A better knowledge of Swedish on the part of the former group will most likely have helped them to benefit from the overall similarity between the English and Swedish article systems, reducing or even neutralizing the "organizational" problem -seeing through the systemconfronting the non-Swedish-born students in some home-language groups.Further, these students will have studied English as part of their schooling in Sweden in the same way and for the same number of years as their "all-Swedish" peers.They will also have had roughly the same exposure to English outside school as the "all-Swedish" group -unlike many non-Swedish-born students.This, then, in combination with possibly beneficial transfer effects from their L2 Swedish, is basically what is involved in the difference between Swedish-born and non-Swedish-born test takers in the "non-Swedish" group.In other words, there is nothing mysterious or magical, with regard to learning English, about being born in Sweden.
To be sure, due to the nature of transfer phenomena and the existence of other causes of error, the picture of cross-linguistic influence emerges as rather less clear-cut than that of the difference in results between Swedish-born and non-Swedish-born students.As regards the good results of the Swedish-born students -in many cases better than that of the "all-Swedish" group -this will come as no great surprise.In the views and experience of many teachers of English in Sweden, students with immigrant backgrounds and a good command of Swedish (usually born in the country) not infrequently seem to have an edge over monolingual Swedish students learning English without any prior experience of learning a new language.In other words, familiarity with another language apart from one's L1 may in itself facilitate the learning of a third language for students with home languages typologically very dissimilar to English. 30There may be 30 Lindblad & Lindblad (1982:45) note that, in answers to a questionnaire sent to headmasters and directors of study at a number of Swedish schools in the early 1980s, "[i]t is ... emphasized that there is a close connection between achievement [in English] and knowledge of Swedish; pupils with a firm command of Swedish, e.g.immigrant children born in Sweden, have no trouble with English ..." Cf. also Ohlander (2006), on the results of students with foreign backgrounds on the Swedish Scholastic Assessment Test (SweSAT), especially English reading comprehension.For discussion of the view that a multilingual background may promote the learning of a new language, see also Odlin (1989:41), Gass & Selinker (2001:135ff.).general and more specific reasons for this, such as increased metalinguistic awareness (e.g., realizing that languages may be organized in radically different ways), on the one hand, and intuitive insights into the function and structure of specific areas of grammar (like the articles), on the other (cf.Jarvis & Pavlenko 2008:194ff.).
With regard to the discussion of cross-linguistic influence in L2 learning, the early work on the learning of English by Finnish-speaking and Swedish-speaking students in Finland is still relevant.Reviewing this work, Ringbom (1987:144) talks about the "consistent ways in which Finnish learners have been shown to differ from Finland-Swedish learners", providing "clear evidence of how central a place the L1 occupies in the L2-learning process."Further, with regard to the English articles, Ringbom (p. 93) points out that "differences between Finnish and Swedish learners ... can be expected" on the grounds that Swedish has an article system, whereas Finnish does not.And sure enough, there is ample evidence, in a number of studies, that this is actually the case, particularly early on in the learning process (pp. 94f.; cf. p. 63).
In my view, the results and conclusions of studies like those accounted for by Ringbom (1987) are very much in line with the gist of the discussion of crosslinguistic influence in the present study.The difficulty of grasping the English article system for Finnish-speaking-students may be seen as comparable to that of many non-Swedish-born students in Sweden, with home languages that are at least as typologically distant from English as Finnish.In addition, many of these students may be assumed to be at an earlier stage of learning English than their Swedish-born peers, and so more sensitive to L1 transfer.The role of Swedish as a source of interlanguage transfer is also highlighted by the Finnish studies, where the similarity between the English and Swedish article systems was seen to provide positive transfer for the Swedish-speaking students. 31 Thus, from a general cross-linguistic perspective, the differences found in the present study between Swedish-born and non-Swedish-born students with immigrant backgrounds are hardly unexpected, although certainly of interest.However, as will also have appeared, the precise nature of such influence in concrete learning contexts, involving a variety of L1-L2(-L3) combinations, as well as a host of other factors and dimensions, is far from clear.It is easy, therefore, to agree with Odlin (2003:442) that "cross-linguistic influence can surface in a wide variety of ways", and that "[a]ccordingly, generalizations about the importance of transfer with regard to a particular structure or, even more, an entire subsystem seem risky at best."For one thing, the interplay between crosslinguistic factors and those of a more universal, developmental nature remains to be fully explored.Under what circumstances, at what levels and in what linguistic 31 Cf.Jarvis & Pavlenko (2008:181), commenting on the wider implications of Ringbom's (1987) findings concerning the role of perceived similarities between English, Swedish and Finnish: "These findings suggest that Swedish and English meet a certain overall-similarity criterion in the minds of the learners ... , whereas Finnish and English do not.These findings also suggest that learners' perceptions of the overall degree of similarity between languages may have an even more profound effect on certain types of transfer than does their perception of the specific, discrete similarities across languages."domains, and how, does cross-linguistic influence make itself felt most strongly?The question of transferability continues to loom large. 32 Finally, from a pedagogical perspective, it should be obvious that students with different home languages have different starting points when approaching a specific foreign language.However, even if the learning process may differ depending on one's L1, this does not mean that the "end product" will also differ as a result of different first languages.What it does mean, rather, is that language teachers, in Sweden as well as many other countries, need to be aware of the cross-linguistic forces at work in multilingual settings and also -in the best of all possible worlds -of the main typological dimensions along which languages may differ, giving rise to positive as well as negative transfer effects.In a global world, where large-scale migration is a prominent feature, such "multicontrastive awareness" should be considered a sine qua non for language teaching in today's multilingual classrooms.

Table 2 :
Overall results of some specific groups

Table 3 .
Overall results of "non-Swedish" home-language groups in comparison with "all-

Table 4 .
Overall results of Swedish-born versus non-Swedish-born students in home-language groups

Table 5 .
Results of some specific groups on written exercises (max.number of points: 41)

Table 6 .
Results of "non-Swedish" home-language groups in comparison with "all-Swedish" group on written exercises (max.number of points: 41)

Table 7 .
Results of Swedish-born versus non-Swedish-born students in home-language groups on written exercises (max.number of points: 41)

Table 8 .
Results of Swedish-born versus non-Swedish-born students in home-language groups on each written exercise, compared to "all-Swedish" group Note: maximum number of points per exercise within brackets; altogether 41 points.