TWO AT THE TIME : THE PARALLEL LEARNING OF FRENCH AND ENGLISH AS FOREIGN LANGUAGES

In this paper, we report on the findings of a research project on the parallel learning of two foreign languages, French and English, by the same pupils in secondary education in Flanders (Belgium). The Flemish education system offers a unique context whereby the same learners acquire two foreign languages simultaneously in similar classroom environments. Not only does this reduce the number of intervening variables (e.g. memory capacity and age) substantially, it also offers an exclusive insight into the interaction between linguistic proficiency, curricular context and socio-psychological dispositions. Some 125 Dutch-speaking pupils, enrolled in the final year of secondary school, were selected from five schools in Flanders. Proficiency in both languages was measured by a cross-linguistically comparable test battery that evaluated productive de clercq and simoens: paralell learning 10 skills (speaking and writing). In addition, the pupils’ attitudinal-motivational dispositions towards these languages were determined by means of a written questionnaire, consisting of some 100 statements that were based upon Gardner’s (1985) AMTB and Baker’s (1992) attitude instrument. In spite of their higher curricular exposure to French, the pupils consistently attained higher proficiency levels in English and were overall more favourably disposed towards English than towards French. Although several factors may contribute to this twofoldresult, our study suggests that the extra-curricular exposure to a foreign language is a significant factor with regard to foreign language learning as it influences both the available learning opportunities and the socio-psychological dispositions towards the foreign language. In turn, the combination of additional extra-curricular input and more favourable socio-psychological dispositions somehow compensate for the discrepancy in formal exposure between French and English in Flemish foreign language classrooms.


Introduction
In the last two decades, the ability to communicate in more than one language has become a key postulate of modern educational theory.Therefore most national educational systems in Europe provide the opportunity for their citizens to learn at least one language other than their own (Bonnet, 2002;EuroStat, 2012).This has resulted in a heightened attention to learning outcomes in foreign language teaching and learning, and in renewed efforts to identify the most effective approaches to foreign language education (e.g.Bonnet, 2002;García, 2009;Rifkin, 2005;Rolstad et al., 2005).To this respect, the current article is concerned with the outcomes and effectiveness of traditional foreign language education in secondary education where the target language (TL) is taught as a separate subject, typically for a limited number of lessons per week, and where the language classroom is usually the major if not sole source of contact with the TL.
Although traditional foreign language instruction is still the most prevalent approach to foreign language education (C. Baker, 2006), the general consensus among educational policy makers, language educators and applied linguistics researchers appears to be that traditional foreign language teaching (FLT) is inadequate, and often woefully so, for developing high levels of functional proficiency in the target language, or is at least significantly less efficient than immersion, CLIL or other content-based approaches to foreign language education (Dalton-Puffer, 2008;Genesee, 1985;Genesee & Jared, 2008;Haunold, 2006;Marsh, 2002;Reagan, 2002;Wesche, 2002).However, the often claimed inferiority of traditional foreign language teaching to immersion or CLIL is still largely an empirical question due in part to the failure of many comparative studies to clac 63/2015, 9-41 de clercq and simoens: paralell learning 12 control for key variables, such as amount of curricular and extra-curricular exposure to the TL, learner background variables, linguistic variables and proficiency measures used (cf.Bruton, 2012;Rolstad et al., 2005).Furthermore, although there are some empirical findings supporting the claim that traditional foreign language teaching methods produce lower than expected levels of proficiency in the target language or fail to meet their own stated goals (Carroll, 1967;Johnstone, 2000;McPake et al., 1999;Reagan, 2002;Rifkin, 2005;Robin, 2000) other research has suggested that, all other things being equal, the outcomes of traditional FLT in primary and secondary education "can be as successful, if not more successful than those of some other instances of bilingual education" (Housen et al., 2011, p. 22).
Not only are the outcomes of the available body of research on foreign language education inconclusive, traditional FLT has also been under-researched.Bruton (2013) claims in this respect that it is the interest in immersion education in the past decade that has "diverted much of the attention away from the so-called mainstream school FL teaching" (p.595).Lately, however, a renewed research interest in FLT has been noted, particularly in FLT at the primary school level in Europe (e.g.Enever, 2011;FLiPP, 2012).The case study of French and English Foreign language education in Flanders (Belgium) presented in this article is part of this renewed interest.But in contrast to most other recent studies, this study focuses on the effectiveness and outcomes of FLT at the end of compulsory secondary schooling.
As already alluded to above, research on FL education is fraught with challenges that may detract from the comparability, representativity and generalizability of its findings.
Many factors, both curricular and extra-curricular, interact in intricate ways in determining outcomes in education, and FL education is no exception.A context where the same learners acquire two (or more) L2s more or less simultaneously may provide better opportunities for investigating theoretical issues in FLT, particularly if the two languages are acquired in similar language learning contexts.With such learners the number of intervening variables may be reduced considerably as the simultaneous L2 learner is assumed to have the same previous knowledge and aptitude, and is always at the same level of socio-cognitive development for learning both languages.The learning process is, however, also influenced by other, linguistic and contextual, factors, such as the target languages, the curricular and the extra-curricular learning context (see also clac 63/2015, 9-41 de clercq and simoens: paralell learning 13 Section 2.2).Also affective variables, such as attitude, anxiety, and motivation, have been shown to be at least as important as language aptitude for predicting L2 achievement (Dörnyei, 2003).
The specific case under investigation in this article draws on a larger study of uppersecondary students learning English and French more or less simultaneously as determined by the curriculum in Flanders, Belgium.It seeks to examine the triangular relationship between (a) the linguistic outcomes of parallel foreign language learning, (b) the socio-psychological dispositions towards these languages, and (c) the language learning context.In what follows, we first present the background to the study by introducing the foreign language education context specific to Flanders (Belgium) and the theoretical framework in which this study takes place.We then outline the methodology for investigating productive language skills and socio-psychological dispositions.The last section discusses and relates the language learning outcomes, the learning contexts, and the attitudinal and motivational dispositions.

The Flemish foreign language education context
The Flemish foreign language context is selected because it provides a setting in which the parallel learning of two foreign languages, French and English, by the same learners can be evaluated.With a few notable exceptions, foreign language instruction in Belgium is started in the 5th grade of primary school (age 10).In Wallonia, this can be English, Dutch or German; in Flanders, pupils obligatorily have French, the other national language, as their first foreign school language.The second foreign language, which is always English in Flanders, is introduced in the first year of secondary school (age 12).Even though Flemish pupils start learning French at an earlier age than English, the efficiency of this head start on learning outcomes has been questioned (Spoelders, 1997).Furthermore, the courses are not taught by specialized foreign language teachers, but by the pupils' general primary school teacher.
The first foreign language is taught for a minimum of two and a maximum of five hours a week, depending on the year of study and the specialization chosen.The second clac 63/2015, 9-41 de clercq and simoens: paralell learning 14 foreign language is taught for minimally one hour and maximally four hours a week.By the end of secondary school (age 18), Flemish pupils in general secondary education will thus have had an accumulated number of roughly 930 classroom contact hours with French as opposed to some 540 hours with English.It should be noted, however, that these are average numbers that may differ considerably depending on specialization and elective courses chosen, though the proportional difference between French and English remains.
Despite the clear discrepancy in the amount of instruction in French and English provided, which would lead one to predict more advanced levels of proficiency for French than for English, the curricula for both languages are remarkably similar in terms of content, structure and teaching methodologies.Moreover, stated objectives and final achievement levels at the end of secondary education are identical for both languages (www.ond.vlaanderen.be).No official explanation for this apparent paradox is given but it is probably prompted by the assumption that the difference in curricular contact between French and English in Flemish schools is compensated by the considerable amounts of additional, extra-curricular contact which Flemish children have with English.As in many parts of the world, English is the all-pervasive language of youth and pop culture, mass entertainment and media.Such additional extracurricular exposure is not assumed to hold for French in Flanders, despite its status as a national language and the native language of nearly 40% of the Belgian population.A recent survey by the European Commission confirms this discrepancy: Flemish students are exposed through different media "at least once a month" to English, but not to French (EuroStat, 2012, p. 105).To this end, this study sought to answer whether there was any empirical justification for these assumptions by assessing the levels of speaking and writing proficiency attained in the two target languages as well as learners' sociopsychological dispositions towards these languages.

The language learning context
As Dörnyei (2003) notes, language learning is a "social event" which thus takes place in a learning context.Ellis (1994) defines the learning context as "the different settings in which L2 learning can take place."While contextual factors do not figure in every SLA theory, the relationship between context and learning has been investigated frequently clac 63/2015, 9-41 de clercq and simoens: paralell learning 15 because the former has been observed to significantly influence linguistic outcomes of foreign language learning (Collentine & Freed, 2004;Ellis, 2008).Learning contexts have typically been characterized through distinctions such as second and foreign language learning, or naturalistic and classroom exposure, in the assumption that different types of learning and different learning outcomes underpin these dichotomies (Housen et al., 2011).Collentine and Freed (2004) noted, however, that not a single learning context provides consistently higher benefits for all dimensions of language learning.Yashima and Zenuk-Nishide (2008) for instance, showed that it is possible for learners in different learning contexts, in their case stay-home learners and study abroad groups, to attain similar outcomes in terms of proficiency, attitudes and communicative behavior.Similar findings were reported by Collentine (2004), who found more beneficial effects in terms of grammatical and lexical abilities in stay-home contexts than in study abroad contexts, while the latter displayed higher narrative capabilities than the former.
Rather than a clear-cut notion, the learning context is comprised of numerous variables, such as the classroom language or the status of the learner's L1 and L2 in the wider society (Colin Baker, 2011).Housen et al. (2011) further define learning contexts as consisting of variables at an individual learning level, a curricular and an extracurricular level.The individual learning context is shaped by "the learners' individual needs, orientations, preferences, abilities, knowledge, personality traits, and their social networks and discourse-interactional practices" (Housen et al., 2011, p. 86).The curricular context comprises more strictly educational variables, from classroom practices, the school's ethos and language policy, to the educational policy dictated by educational authorities.At this level, we equally find what Gardner (2001, p. 79) refers to as the formal learning context, i.e. "any situation in which [formal] language instruction takes place".Finally, the extra-curricular context consists of "the wider sociolinguistic, demographic, cultural and institutional conditions […] that are somehow beyond the direct control of curricular intervention" (Housen et al., 2011, p. 87).Learning at this level can equally occur in the informal learning context, that is, a setting "where the intent is not instruction in the second language but rather exposure to it for some other purpose" (Gardner, 1985, p. 148).clac 63/2015, 9-41 de clercq and simoens: paralell learning 16 What these contextual levels have in common is that they govern input and output conditions, which are essential in the development of a FL.According to Ellis (2008), learners not only benefit from interaction with native speakers, but also from exchanges between L2 learners.These interactions are especially important in contexts where the TL is learned in foreign language contexts.Individual variables may then influence the degree to which a learner actively seeks out TL input, while curricular and extracurricular factors influence the available input and output opportunities for the learner.
The three levels thus interact to shape a dynamic learning context.
The contextual configuration at these three levels may not only differ between different learning populations, but also within contexts that are superficially similar.In the Flemish situation, for instance, there are some important differences in the curricular exposure to French and English, although the outcomes are expected to be similar (see previous section).At the same time, we may expect contextual differences at the extracurricular level, such as input and output conditions, the status and functional roles of the languages (Housen et al., 2011), and at the individual level, such as attitudes and motivations.
The learning context, then, is a significant factor in the development of the TL and it has been found to influence both linguistic and non-linguistic outcomes of FLE (Yashima & Zenuk-Nishide, 2008).Under linguistic outcomes we understand the learner's oral and written proficiency, both in comprehension and production, as well as his metalinguistic and sociolinguistic awareness.In the Flemish educational system, these linguistic outcomes have been explicitly formulated as final attainment levels for general secondary education (AKOV, 2014).The relation between context and linguistic proficiency has been elaborately studied in SLA research (see Ellis, 2008 for an overview).Contextual effects are usually tested to determine the effects of, amongst others, age (Birdsong, 2006;DeKeyser, 2013), study-abroad (Collentine & Freed, 2004) and content-based learning (Dalton-Puffer, 2008).In the case of foreign language learning, there has been a surge in research on immersion-based approaches, whereby learners are taught non-language courses in a FL.If these approaches, such as CLIL, are promising to some (Van de Craen et al., 2007), others are more skeptical and call for more rigorous research (Bruton, 2011a(Bruton, , 2011b)), as well as attention to the role of the national context (Sylvén, 2013).Bruton (2013) furthermore hopes to at least partly clac 63/2015, 9-41 de clercq and simoens: paralell learning 17 refocus the SLA research agenda to traditional mainstream FLL, since this is still the most important type of language learning.Apart from proficiency-related outcomes, the attainment levels set by the Flemish educational authorities equally mention non-linguistic outcomes in the form of the learners' socio-psychological dispositions towards the TLs.The motivational dimension plays an important role in FLL, because regardless of the numerous advantages of learning other languages, "they are not absolutely necessary" (Gardner, 2007, p. 10).
The exact role and functioning of L2 learning motivation is, however, still subject to debate and has been described in a plentitude of theoretical models (see, for instance, Dörnyei, 2003).In Gardner's (1985) socio-educational model, motivation is supported by integrativeness and attitudes toward the learning situation.Integrativeness points to a "positive interpersonal/affective disposition toward the L2 group and the desire to interact with and even become similar to valued members of that community" (Dörnyei, 2003, p. 5) and is shaped by the extra-curricular context.Attitudes towards the learning situation, on the other hand, deal with "the individual's reaction to anything associated with the immediate context in which the language is taught" (Masgoret & Gardner, 2003, p. 127) and is more influenced by the curricular context.Integrativeness and attitudes towards the learning situation are, then, causally linked to motivation, which in turn relates strongly to L2 achievement (Gardner, 2007;Masgoret & Gardner, 2003;Tremblay & Gardner, 1995).
Other affective variables relevant to L2 learning are language anxiety and related concepts such as Willingness To Communicate (WTC).Language anxiety is often considered separately from attitudes and motivation, because it is more strongly linked to emotion in the sense that it involves an automatic and physiological reaction and is related to fear (MacIntyre, 2002).Whereas L2 motivation facilitates language learning, language anxiety may have a debilitating effect on the learning outcomes (S. C. Baker & MacIntyre, 2000;Gardner, 1985;MacIntyre & Gardner, 1994).
It transpires from the previous discussion that the relationship between context, sociopsychological dispositions and linguistic proficiency is multidirectional.High motivation may lead a learner to seek out more extra-curricular exposure and thus shape his learning context differently, while input and output conditions may equally influence the learner's motivation.Similarly, highly proficient learners may be more motivated to clac 63/2015, 9-41 de clercq and simoens: paralell learning 18 look for learning opportunities.In this paper, we examine the triangular relation between the language learning context, linguistic outcomes (i.e.quantitative proficiency) and non-linguistic outcomes (i.e.attitudes and motivations) in the frame of the more or less simultaneous learning of French and English through mainstream foreign language learning.

Research questions
The following research questions are formulated: 1. What are the levels of speaking and writing proficiency for French and English and in what ways do the levels for French differ from those for English?
2. What are the socio-psychological dispositions towards French and English and in what ways do these differ for French and English? 3. How do proficiency scores and socio-psychological dispositions within languages relate?
Since the purpose of this study is explorative, we hypothesize that the students' proficiency levels (H1) and their socio-psychological dispositions (H2) will be similar for French and English.These hypotheses are moreover supported by the identical official final attainment levels set for French and English.Finally, we expect to find a positive correlation between the students' socio-psychological dispositions and their proficiency levels (H3).

Design
The study presented here is part of a larger research project at the Vrije Universiteit Brussel, Belgium.For this project, a total of 125 students, aged 17-18, enrolled in the final year of secondary school, was selected from five secondary schools across Flanders.Two schools (C and D) were located in the province of Antwerp, two in Eastclac 63/2015, 9-41 de clercq and simoens: paralell learning 19 Flanders (A and E) and one in Limburg (B).Care was taken that none of the schools fell within the immediate Francophone sphere of influence, i.e. the officially bilingual yet predominantly French-speaking capital of Brussels, and the Dutch-French border that separates Dutch-speaking Flanders from Francophone Wallonia.Furthermore, all pupils that participated in the survey were enrolled in Modern Languages 1 , which is the study programme offering the highest exposure to foreign languages.The choice of this programme allowed us to determine the maximum level of linguistic proficiency within foreign language education in Flanders.
Whereas this study focuses on the outcomes of FLL in terms of production skills and socio-psychological dispositions, the larger research project examined other factors, such as receptive language skills, metalinguistic knowledge, classroom interaction and textbook composition.A more detailed description of the original study can be found in Housen et al. (2003).

Productive skills
Productive skills were assessed for writing and speaking by using two narratives in the form of picture-description tasks.Both tests were administered under the same conditions, following identical procedures, namely in a classroom context during regular French and English classes.The French and English tests were conducted separately, with at least a week's difference to avoid external factors such as test fatigue and test familiarity so as not to distort the results.The most important condition for the tasks to meet was cross-linguistic equivalence of measurements to ensure comparability of the evaluation of proficiency in English and French.The tasks were designed to evaluate the same aspects of proficiency with equal levels of difficulty in the two foreign languages.To this end, the validity of the tasks was evaluated by a panel of 1 Students in Flanders are required to enrol in a specialization programme, which determines the curricular weight of some courses.In A-stream education, some of the options are economics, to science and modern languages.clac 63/2015, 9-41 de clercq and simoens: paralell learning 20 independent specialists, consisting of FL teachers, members of the pedagogical inspection and bilingual (French-English) linguists specialized in developing tasks.
The written narrative involved three photographs presenting a person getting in a car, a traffic jam and a traffic accident.The participants were instructed to write a 10-15 line story linking these images.For the oral narrative, the participants were asked to retell the Frog story (Mayer, 1969), an instrument that had been employed by many previous studies on first and second language development (cf.Berman & Slobin, 1994).The students retold the story in individual fifteen-minute interviews with near-native speakers of French and English.The interviews were then tape-recorded and transcribed in CHAT format (MacWhinney, 2000).
This article draws on a subsample of 50 students, of whom all the relevant data, i.e. a written and spoken narrative, as well as a fully completed questionnaire, were available.
To ensure an equal distribution of students among the five schools involved, 10 students per school were selected.
The students' productions were analysed in the Complexity, Accuracy and Fluency (CAF) framework, using complexity and accuracy measures 2 .Although interesting, fluency measures were not included because they are difficult to operationalize uniformly in written and oral texts.Linguistic complexity was operationalized for the lexicon using VocD (McCarthy & Jarvis, 2007) in CLAN (MacWhinney, 2013).Syntactic complexity, another dimension of linguistic complexity, was operationalized as the Subclause Ratio (SubR) and the Mean Length of Clause (MLC) (Wolfe-Quintero et al., 1998).The subclause ratio calculates the proportion of subclauses to main clauses in a text, while the mean length of clause calculates the average number of words per clause.
Clauses were defined using a linguistic definition that distinguishes between main clauses, subclauses, minimal clauses and incomplete clause.Clauses were considered as main clauses if they were syntactically independent and not embedded in another clause.Subclauses, then, are clauses which are part of, or grammatically dependent on another clause.A clause was treated as a minimal clause if it was pragmatically 2 See Ellis and Barkhuizen (2005) and Bulté and Housen (2012) for a discussion of these measures.clac 63/2015, 9-41 de clercq and simoens: paralell learning 21 appropriate but lacked one or more explicit grammatical constituents.In cases where an absent constituent rendered the clause ungrammatical, it was marked as an incomplete clause.
Finally, the percentage of error-free clauses was calculated as a general accuracy measure.Accuracy was assessed by two researchers at the VUB, who were native or near-native speakers of French and English.

Socio-psychological dispositions
The participants' socio-psychological dispositions were assessed through a questionnaire containing approximately 125 statements on 5-point Likert scale, based on the AMTB (Gardner, 1985).These items are concerned with topics ranging from the students' motivations for learning French and English to their appreciation of the languages and the languages courses.Items that related specifically to French and English were always balanced (e.g.I like speaking French/I like speaking English).The items in the questionnaire not related to French and English were not included in this analysis.
Two separate Promax-rotated Principal Component Analyses (PCA) were carried out on the data to verify whether certain constructs underlay the participants' answers and whether these constructs were similar for French and English.In total, 34 items per language, recoded where necessary, were included in each separate PCA.For both languages, the items can be grouped into four comparable components, namely the attitudes towards the culture associated with the language, attitudes towards the instructional setting, motivations for learning the languages and attitudes towards speaking the languages.The results of the PCA will be discussed in more detail in Section 4.2.
The advantage of using these components is that they provide an alternative to raw scores.The interpretation of the Likert scale can be highly subjective and raw scores may thus be misleading.A disadvantage to the components is that they provide no information about what the attitudes actually are.For this reason, we will also provide an indicative average for the scores on every component.These averages were calculated by averaging the scores of the seven items that contributed most strongly to clac 63/2015, 9-41 de clercq and simoens: paralell learning 22 each component.This score is only intended as an approximation and interpretative help.

Linguistic proficiency
Table 1 presents the descriptive statistics for the measures of linguistic proficiency.A couple of preliminary remarks are warranted.First, score differences do not necessarily reflect higher or lower proficiency, since some measures are more sensitive in one language than in the other.The percentage of error-free clauses is, for instance, a sensitive measure for French since mistakes on article gender are easily made repeatedly and may affect the score significantly.Secondly, we have not examined native speaker benchmark data, although it is possible that linguistic preferences differ from French to English.For instance, it is possible that native speakers of French and English use subordination to a different extent.In this respect, complexity measures are not always linearly related to proficiency.We will proceed with a brief discussion of the differences between French and English, first for oral proficiency, then for written production.In the second part, we will examine whether scores for one language or mode correlate with those for another language or mode.
English vs. French: Speaking The mean values for lexical diversity, measured through VocD, were very close for English (35,76) and French (33,12), but are nonetheless significantly higher for English (p < 0,05).The standard deviation for VocD is slightly higher in French (st. dev. = 8,38) than in English (st.dev.= 5,80).Although the difference between the mean scores is minimal, it is again statistically significant (p < 0,05).
This suggests a similar distribution around the mean, although it also indicates that the amount of subordination in the texts is highly variable in both French and English.
The mean length of clause is again similar in English (mean = 5,99) and French (mean = 5,74).In both languages, a clause contains slightly less than 6 words on average.The slight difference is significant nonetheless (p < 0,05).
The largest difference, however, is to be found in the percentage of error-free clauses, which is significantly higher (p < 0,05) in English (mean = 0,84) than in French (mean = 0,62).Equally striking is the higher standard deviation in French (0,14) when compared to English (0,07).The lower mean is unsurprising considering the sensitivity of the measure in French.Whether this also explains the higher variation is unlikely.
Finally, the scores for English were all normally distributed, whereas this is not the case for French.This indicates that the pupils form a more heterogeneous group with regard to English proficiency than in French.
In sum, the scores for English were higher for all the measures, which points to a higher degree of oral proficiency in English.Nonetheless, the differences are not as pronounced for every measure.That is, while the scores for subordination and error-free clause were notably higher in English, the difference between French and English with respect to lexical diversity and MLC is marginal.In this case the different scores for the clac 63/2015, 9-41 de clercq and simoens: paralell learning 24 two syntactic complexity measures indicate that these measure are (more or less) independent constructs that do not necessarily develop in the same way (cf.infra).

English vs. French: Writing
In written production, the average scores for lexical diversity are again fairly close for English (D = 68,59) and French (D = 70,01).In this case the standard deviation indicates that lexical diversity varies more in English (D = 26,60) than in French (D = 18,62).The differences that were observed here are not significant (p = 0,37).
The scores for English do not follow a normal distribution, which indicates some level of heterogeneity.For French, the scores are more homogeneous, as the scores for subordination ratio and the MLC follow a normal distribution.
In sum, the differences in written production are only significant for the subordination ratio and the percentage of error-free clauses.Where the differences are not significant, the average scores are even slightly higher for French than for English.

English vs. French: correlations
In this section we will discuss the relation between speaking and writing skills across languages.Table 2 presents the observed correlations between scores for French and English.The scores of speaking proficiency will be addressed first.Here, the only correlations between French and English were for lexical richness (r(50) = 0,505, p < 0,001).In other words, speakers whose English oral narrative contains a rich vocabulary will also have a rich vocabulary in their narratives in French.
clac 63/2015, 9-41 de clercq and simoens: paralell learning 25 For the writing skills, only the scores for lexical richness correlated positively (r(50) = 0,331, p < 0,05).No evidence is found for a relation between the use of subordination, the mean length of clause and error-free clauses in French and English.The absence of correlations between French and English for writing and speaking scores may have two explanations.Either the underlying constructs function independently in the two languages, or the data does not exhibit enough variation to determine correlations.The second case may be true for the absence of (strong) correlations in MLC.We have already mentioned that the mean MLC for writing and speaking is close to six words in French and English.The standard deviation for the oral narratives was low, at around 0,5 for English and French.For the written task, the standard deviation is slightly higher, at around 0,7 for English and 0,9 for French.These relatively low values may explain why no correlation was found for the oral narratives and only a low correlation for the written narratives, although the mean values suggest a clear correspondence.A similar reasoning may be applied to the scores for accuracy.The limited variation for English, around 7%, may prevent any significant correlations to surface.

Interim comparison: Writing vs. Speaking
The results indicate that the English and French scores differ more for some measures than for others, regardless of the mode of production.Most notably, the scores for the subordination ratio and the percentage of error-free clauses are higher for English in both written and oral production.While differences of lexical diversity are significant in oral production, the actual means reveal that this difference is only marginal.In other clac 63/2015, 9-41 de clercq and simoens: paralell learning 26 words, on average, the pupils do not produce noticeably longer clauses, or introduce more different words in one language compared to the other.
With regard to the distribution of the scores, we observed less variation in the scores for English in oral production than in the written mode, while the opposite is true for French.This suggests that, in some cases, statistically significant differences are the result of different distributions rather than different average scores.
In all cases, the language that was used in the written stories was more complex and accurate than in the oral narratives.The only scores that remained more or less stable across modes was mean length of utterance.
Turning to correlations between speaking and writing skills (Table 3), only one correlation can be observed for English, i.e. between vocD scores for writing and speaking (r(50) = 0,330, p < 0,05).While this correlation may be unsurprising in that speakers can be assumed to have the same vocabulary at their disposal in both modes of production, this is not a guarantee that they will also rely on it in the same way.The lack of evidence for other correlations indicates a disparity between written and spoken proficiency for English.For French, most scores correlate significantly across modes.We find a positive correlation for the scores for subordination (r(50) = 0,339, p < 0,05), MLC (r(50) = 0,412, p < 0,05) and error-free clauses (r(50) = 0,331, p < 0,05), but surprisingly not for VocD.Though the correlations are always average, they indicate that proficiency in oral production is significantly related to proficiency in written production, an observation that could not be made for English.

Socio-psychological dispositions: results
This section reveals the results from the Principal Component Analysis (PCA).Table 4 presents the four components that were distinguished.The descriptions assigned to the different components are, of course, subjective approximations, though we believe them to be adequate representations of their underlying constructs.The four components explained 50.97% of variance for the French items and 50.76% for the English items.

Component Examples
Attitudes towards culture I learn French/English to read literature; I prefer French/English content in the original language.
Anxiety French/English is a difficult language; I'm nervous when speaking French/English; other pupils are better at French/English.
Attitudes towards instruction I like learning French/English; the French/English courses emphasize writing; I often answer questions during the French/English courses.
Motivation I learn French/English to travel, make friends; listen to the radio; for my job later.
Table 4 Four components extracted from the PCA Since a PCA only provides information on correlations between items and not on the raw scores, we also provide approximate averages (Table 5).As mentioned previously, these scores were calculated by averaging the scores for the seven items that contribute most to each component.They are merely intended as interpretative guides to the PCA.
Scores under 3 signal negative attitudes, while scores above 3 signal positive attitudes.
The The different types of attitudes towards English all significantly correlate, with one exception (Table 7).Attitudes towards culture and attitudes towards instruction only show a trend toward correlation (r(50) = 0,277, p = 0,051).
Note that on average these correlations tend to be low to average.In other words, the components are still evaluated as separate items.The highest correlation can be observed for culture (r(50) = 0,655, p < 0,001), then for motivation (r(50) = 0,521, p < 0,001), indicating similar attitudes for French and English towards the culture associated with the foreign language and an interest in taking part in authentic activities in that language.For both languages, the attitudes towards culture and the motivation are positive.
Likewise, attitudes towards French and English instruction tend to be positive.
However, the correlation between French and English instruction is slightly lower (r(50) = 0,325, p = 0,021) but still significant.There is thus only a moderate overlap between the attitudes toward French and English instruction.
The only non-significant correlation that was observed was between anxiety in French and English (r(50) = 0,205, p = 0,153).This is not surprising since scores for anxiety did not correlate with any of the other attitudes towards French.

Socio-psychological dispositions and linguistic proficiency: results
Finally, Tables 9 and 10 present the correlation analysis of the students' sociopsychological dispositions and their language proficiency.A first observation from these tables is that no clear picture emerges from the data.Only 8 out of 64 potential correlations were significant (p < 0,05).There seem to be no consistent differences or similarities across speaking and writing scores.The scores for lexical diversity correlated most frequently with English Motivation (r(50) = -0,389, p < 0,05), English Culture (r(50) = 0,309) and French Culture (r(50) = 0,331, p < 0,05) for writing and with French Integration for speaking.Remarkably, two of these correlations are clac 63/2015, 9-41 de clercq and simoens: paralell learning 30 negative.Except for English Speaking, all scores for the socio-psychological dispositions correlate with at least one measure of complexity or accuracy.We hypothesized that linguistic outcomes for French and English at the end of secondary school would be similar (H1).Our analysis of the students' oral and written productions, however, has provided no proof for this hypothesis.To the contrary, the data even indicate higher proficiency levels for English than for French.However, a fine-grained analysis of the different components of linguistic proficiency exposed clear differences between the two languages for both speaking and writing.
Firstly, the overall speaking scores are higher for English than for French.This finding could not be predicted on the basis of the students' exposure to the languages in the A second observation, however, points out that the students' writing skills were more similar across languages than their speaking skills.In the case of lexical diversity (VocD) and syntactic complexity (MLC), scores for French were even higher than for English in written production.These findings suggest that writing is more dependent clac 63/2015, 9-41 de clercq and simoens: paralell learning 32 upon the curricular context than speaking is.Arguably, formal school settings offer pupils essential opportunities to practise writing skills that may not be present to the same extent in extra-curricular settings.
Finally, the effect of the wider, out-of-school context also shows in the relation between written and oral proficiency for each language.In French, speaking scores significantly correlated with writing scores.This consistency is presumably the result of the homogeneous, curricular context on which French learning predominantly depends.In turn, the lack of correlations between written and spoken proficiency for English hints at a strong extra-curricular, and thus heterogeneous, influence on the language learning process.
We have presented the results of two pairs of tasks, written and oral, which were completed by the same students in two foreign languages.Even though the students had received more hours of instruction for French than for English, their proficiency levels were generally higher for English than for French.These results call for a fine-grained approach to context as a multi-layered construct in order to account for outcomes of traditional foreign language education.The data has highlighted that, even for superficially similar FLL contexts, the effectiveness of the language learning process is significantly determined by non-curricular factors.Although these factors fell outside the scope of this study, explanations for the differences in proficiency could be found, for instance, in the input and output conditions provided by the extra-curricular context.

Socio-psychological dispositions
In line with the postulated expectations (H2), the results of the survey reveal that students tend to have generally positive attitudes towards French and English, yet these are expressed more strongly for English.One exception to this is that participants displayed more anxiety to speak French than English.Rather than signalling a reluctance to speak French, we believe the discrepancy between English and French scores indicates a greater confidence towards using English.This may equally be explained by the different status of English and French.The pervasiveness of English in media and on the internet yields to a higher exposure to English in non-instructional settings.Moreover, this exposure is not necessarily to L1 English because of the language's status as a lingua franca, which sets the bar of linguistic achievement clac 63/2015, 9-41 de clercq and simoens: paralell learning 33 considerably low.It is clear that these extra-curricular factors may reduce students' anxiety towards their 'willingness to communicate' in English (S. C. Baker & MacIntyre, 2000).Exposure to French, on the other hand, is mostly limited to the classroom.It is thus not surprising that students, whose main exposure to French is in a context where their proficiency is graded through tests and other classroom practices, experience greater anxiety towards communicating in the language.Also, in Belgium and its surrounding countries French is predominantly used within a native speaker community, not as a lingua franca spoken by people whose proficiency is less than perfect.In sum, a low to non-existing extra-curricular exposure to French resulted in high levels of L2 anxiety, while the opposite is true for English.
Nonetheless, anxiety in French works independently of the pupils' attitudes towards the French-speaking culture, towards French courses and towards the motivation to learn French as these three components prove to correlate.It follows that, for example, students that are highly motivated to learn French, also tend to have positive attitudes towards French instruction and towards the French-speaking culture.However, despite positive attitudes towards language instruction and the culture of the language and despite a high motivation, the learners may still experience great anxiety towards speaking French.In contrast, anxiety to use English runs parallel with attitudes towards English instruction, towards English culture and with the motivation to learn English.
These findings correspond to what Baker & MacIntyre (2000) concluded about the strong dependency of L2 anxiety on the learning context.
A correlation analysis of the components across languages echoes the previous findings: attitudes towards language instruction, towards the culture of the language and towards the motivation to learn the language correlate between French and English, with the exception of anxiety.This indicates that, for instance, pupils that are highly motivated to learn French and that are positively disposed towards French culture and French instruction are also highly motivated to learn English and positively disposed towards English culture and English instruction.However, it does not imply that pupils that are confident speaking English are also confident speaking French.
To conclude, results showed that, in general, Flemish pupils are positively disposed towards foreign languages.When the two languages in question are compared, however, the overall attitudes and motivations are consistently higher for English than for French.
clac 63/2015, 9-41 de clercq and simoens: paralell learning 34 With respect to anxiety, both languages tend to deviate, as the students feel confident towards speaking English, but rather anxious towards using French.Once again, the presence of English in the wider, extra-curricular context emerges as the most likely explanation for these observations.

Socio-psychological dispositions and linguistic proficiency
The comparison of L2 motivation and the CAF scores is more problematic and refuses a consistent explanation.It is, for instance, unclear how the identification with the L2 community relates to lower lexical diversity scores in French and English.Nor is it immediately apparent why there are no correspondences between the written and spoken productions in terms of correlations with L2 motivation.
There may be a number of explanations for the absence of more significant and meaningful correlations.First, the results could imply that the relation between L2 motivation and L2 proficiency is not as strong as presumed.Secondly, it is possible that there was not enough variation in the CAF scores to observe significant correlations.
This possibility was already mentioned previously and could be further investigated by examining a more heterogeneous group of learners.It is, however, more probable that the CAF measures do not accurately reflect the type of linguistic achievement discussed by Gardner (1985).The measures may be too detailed to expect significant relations between proficiency and L2 motivation.Moreover, we have exclusively examined production data, which only represent one type of linguistic proficiency.Ultimately, the motivational constructs analyzed in this study remain macro-level traits of the learner, which may be inefficient in explaining micro-level behaviour as measured by the CAF framework.CAF measures have been used effectively to investigate the effects of task complexity, as demonstrated by the work of Skehan (1998) and Robinson (2007).
In a similar vein, Dörnyei (2003, p. 21) states that "SLA research, naturally, focuses on the development of language knowledge and skills and therefore analyzes various language processes from a micro perspective, which is incompatible with the macro perspective adopted by traditional motivation research."As a consequence, we have not been able to provide evidence for our third hypothesis (H3  (Gardner, 2007).

Conclusion
By analysing the parallel learning process of two foreign languages in traditional foreign language education in Flanders, this study has shown that the outcomes of FLT in terms of proficiency and socio-psychological dispositions cannot be explained exclusively by curricular factors, but that extra-curricular factors play a significant role as well.
Although French language courses in Flanders are generally started two years earlier and have considerably more classroom contact hours than English language courses, the ultimate attainment levels for English proficiency exceeded those obtained for French.
At the same time, Flemish children tend to be more favourably disposed towards English than towards French.While several factors may contribute to this twofold result, both findings are presumably due to the impact of the wider, out-of-school context.If we consider that, despite the significantly lower curricular prominence of English, the students attain higher proficiency levels for English and, equally important, are also more confident when using the language, then the effect of extra-curricular exposure can indeed not be underestimated.The combination of additional extracurricular input and more favourable socio-psychological predispositions thus somehow compensate for the considerable discrepancy in formal exposure between French and English in Flemish foreign language classrooms, at least as far as speaking and writing proficiency are concerned.In this respect, French is considered to be more of foreign language in Flanders than English is (cf.Goethals, 1997).
However, care should be taken when it comes to the implications of the study.Firstly, the results should not be taken to imply that curricular exposure to the FL is inferior to extra-curricular factors of the kind described here.The high levels of English or more effective foreign language education, but for complementary input in the extracurricular context that is engaging for the learners and which promotes the FL's usefulness.
Secondly, the impact of extra-curricular factors is more strongly manifested in speaking proficiency, where the advantage for English is most significant, than in the writing proficiency, where the differences between English and French are less compelling.
This reflects the nature of the extra-curricular and curricular contact that Flemish children have with English and French: Flemish children may hear and speak English outside the classroom much more than they hear and speak French, but they do not necessarily write English any more than they write French.In turn, writing is considered to be a more academic skill as it is practiced more often than speaking in traditional foreign language education.
Thirdly, in conducting this research, we were hampered by the small pool of pupils available since only five schools could be included at the time of the pilot study.It is clear that these findings cannot simply be extrapolated to all foreign language education in Flanders.Therefore, the general picture that emerges is still a tentative one.Despite the small sample size, however, we feel that this study has provided interesting exploratory findings of the relationship between extra-curricular input factors and the development of productive skills in foreign language settings.
To conclude, our study suggests that the amount of extra-curricular exposure to a foreign language is a significant factor with regard to L2 learning as it influences both the available L2 learning opportunities and the socio-psychological dispositions towards the L2.This runs counter to the traditional definition of a foreign language being learned in an environment where, as opposed to second language learning, exposure to the TL is generally absent.We, however, assert that future research should consider a more fine-grained approach to the extra-curricular learning context in investigating the outcomes and effectiveness of foreign language education.This is in line with what (Hymes, 1972) claimed more than four decades ago: "the key to understanding language in context is to start not with language but with context… [and then to] systematically relate the two" (pp.xix-lvii).
clac 63/2015, 9-41 curricular context, nor does it conform to the official final attainment levels decreed by the Ministry of Education.Arguably, factors situated in the extra-curricular context, in the form of input and output conditions but also the status and function of the FLs in the wider society, could have compensated for the lower curricular exposure to English.Since we have not explicitly examined the pupils' extra-curricular exposure to the FLs, further research should clarify whether the effect of extra-curricular contact sufficiently compensates curricular inequalities.
proficiency attained in Flemish education are the result of the combined effect of curricular and extra-curricular input factors.The net-effect of the curricular factors is most clearly seen in the case of French, where development relies almost exclusively on what the pupils pick up from the classroom, but where nonetheless fairly advanced levels of proficiency are attained.Therefore, this study calls not so much for additional clac 63/2015, 9-41 de clercq and simoens: paralell learning 36

Table 6
tableshows that all the averages are positive, except those for Anxiety in French, which are slightly negative to neutral.Correlation analysis of socio-psychological dispositions for French regression scores on the components.Turning to the correlation analysis of socialpsychological dispositions in French (Table6), an interesting pattern can be observed.

Table 7
Correlation analysis of socio-psychological dispositions for English Finally, Table8presents the results of the correlation analysis of component scores in English and French.The table echoes the previous findings: the different components correlate across languages, with the exception of anxiety.

Table 8
Correlation analysis of socio-psychological dispositions in English and French and simoens: paralell learning 35 investigation in the relation between micro-level CAF measures and classroom learning motivation, which is oriented to the task at hand ).A clearer picture might emerge from a comparison of L2 motivation and a more general proficiency score, such as one obtained by a Cloze-test.Alternatively, future research may find a more fruitful clac 63/2015, 9-41 de clercq