Development of productive language skills through language exchange in primary schools in Switzerland – An exploratory intervention study

1 Introduction

Language exchange and mobility are considered effective means of promoting language learning because they enable learners of all ages, and especially younger learners, to experience, use, and acquire language in meaningful social contexts, combining extracurricular and intra-curricular learning (Llanes and Muñoz 2013). While mobility programs are popular throughout Europe, especially among tertiary level students and teenagers (e. g. Erasmus & Youth in Action), and while programs for younger learners also exist, not many learners engage in them. Especially at primary school level, exchange activities are rare (Krüger-Potratz et al. 2018).

Switzerland, which is officially quadrilingual and where different cultural and linguistic communities live together in a small area, appears to be a privileged context for various forms of exchange, also at lower levels of schooling. Most Swiss communities are monolingual due to the territorial principle anchored in the Federal Constitution of the Swiss Confederation (Bundesverfassung der Schweizerischen Eidgenossenschaft 1999), but the country’s internal cohesion depends heavily on a certain degree of individual multilingualism. The Languages Act, adopted in 2007, based on Article 70 of the Federal Constitution, binds the Confederations and the cantons to encourage mutual understanding between the language communities through domestic exchange (Bundesgesetz über die Landessprachen: Art. 14). Language exchanges as an important field of action in the Swiss education system, promoting multilingual and intercultural skills as well as motivation at all school levels, have recently been supported by a national exchange strategy (Schweizerische Eidgenossenschaft and EDK 2017). The current Swiss curricula, such as Lehrplan 21 or Plan d’Etudes Romand, also recommend language encounters and exchange activities and outline corresponding learning goals (Egli Cuenat et al. 2015).

Given this favourable starting position regarding educational policy, it could be assumed that language exchanges across language regions are common in primary schools in Switzerland. However, this is not yet the case. In the school year of 2015/16, when this study was planned, less than three percent of all Swiss pupils in compulsory education (grades 1–9) benefited from an exchange within the country (Bundesamt für Statistik 2017; ch Stiftung 2017). At the primary level, which is the focus of the present study, participation is even lower, at just under 0.5 percent throughout Switzerland. Among the reason for this low participation rate, the additional workload for teachers associated with an exchange seems to be one of the most weighty ones. They are already very busy in their everyday work and, therefore, not always willing to take on additional work (Borel and Gygax 2011). In the case of primary school pupils, the young age and the limited language skills of the pupils make things even more difficult.

In light of this scarcity of exchange activities even in contexts of geographical proximity and the challenges associated with their implementation at lower grades, there is a need to develop age-appropriate forms of exchange, which are targeted at intensifying language learning as well as the need to examine their impact. So far, there are only a few studies available on the impact of language exchange projects in Switzerland (Ogay et al. 2007; ch Stiftung 2006; Heinzmann et al. 2014). This is particularly true at the primary level. In general, exchange projects with this age group have not yet been the subject of much international research either. The project presented in this paper contributes to closing this gap^[1]. The aim was to develop an exchange setting at primary level of shorter duration between partner classes in German-speaking and French-speaking Switzerland which can be implemented relatively easily and to test its effectiveness with an intervention group and a control group. The learners were 6^th grade students. The focus of the study was on curriculum-relevant speaking and writing skills as well as language learning motivation. The development of pupils’ target language skills was only assessed in the German-speaking classes. The present paper focuses exclusively on the effect of the exchange activities on the development of these target language skills.

2 Theoretical and empirical underpinnings

Research on language stays and study abroad has always been strongly interested in their effectiveness for language development. This outcome-oriented research has shown that longer stays (usually in the form of study abroad) have the potential to improve the language skills of those involved on a long-term basis (see e. g. Freed 1995; Engle and Engle 2004; Magnan and Back 2007). Most progress can be found in the areas of fluency (Llanes and Muñoz 2009; Segalowitz and Freed 2004), vocabulary (Dewey 2008; Serrano et al. 2011), and pragmatic skills (Regan 1995). As far as writing is concerned, the picture is less clear, however, as findings are contradictory (Pérez-Vidal and Barquin 2014; Pérez-Vidal and Juan-Garau 2009; Sasaki 2004, Sasaki 2007, Sasaki 2009, and Sasaki 2011).

However, since longer language or study stays in a foreign-language area are mainly undertaken at upper secondary or tertiary level, research with regard to younger learners is relatively rare (Brunner 2015). The larger study by the Franco-German Youth Office (Krüger-Potratz et al. 2018; Melin and Wagner 2015) is worth mentioning. Using ethnographic videos and interviews with pupils and teachers, the study dealt with five-day encounters between school classes from five French and five German primary schools (generally two encounters of which one was in each partner’s country). A specific feature of the exchange situations was that the learners were not taught each other’s language as their first foreign language. The focus was not on foreign language learning, but on informal, social and intercultural learning. The study revealed a wide range of mostly non-verbal communication strategies, forms of intercultural learning as well as changes in self-perception and perception of others that were developed by the children. The authors point out that while exchange activities at primary school level are still scarce, they can pave the way for linguistic and intercultural learning. Based on their findings, they formulate recommendations for the design of exchange situations and teacher training, even in situations with no or very limited target language skills in the partner language.

Similarly, the study by Krok et al. (2010), also funded by the Franco-German Youth Office, documented the learning experiences of children who took part in an international children’s encounter of at least 5 days. Although this study, too, focused on intercultural rather than language learning, it nevertheless provides interesting information on the language use of the pupils. Children from Germany aged between eight and twelve years were examined. The children participating in the group encounter stayed in a shared accommodation during the encounter. The survey of the children showed that they used a variety of communication methods when communication problems arose: they used a different language, communicated with gestures, sought help from counselors, spoke their first language or painted a picture. Another important finding of the study was that experiences and learning effects are variable. Those children, who due to their age, experience and attitudes, had more numerous resources, were more often represented in the group of learners who communicated in a versatile and flexible way – i. e. also used their resources. Older children, who had a wider range of experience and were more likely to have foreign language skills, used more versatile means of communicating with the exchange students. Furthermore, the observations indicated that the experiences and thus also the learning effects differed more among the children the more their own initiative was required when making contact with participants from abroad. This shows the importance of the didactic setting and support of such an encounter.

With her case study of a two-week exchange between two grade 6 classes (13-year-old and younger students) in Ontario and Quebec, MacFarlane (2001) examined the question if and how exchanges and classroom language learning were complementary. Focus groups, observation, and informal teacher interviews were used. Based on Wong Fillmore’s multidimensional Teachability Model, social, linguistic, and cognitive processes were identified and compared between classroom language learning on the one hand and the exchange context on the other hand. The study concludes that these two language settings are complementary. The classroom and the exchange setting provide different L2 input contexts which are both needed for successful language acquisition. Language skills acquired in the classroom do not transfer automatically to authentic interactions with native speakers, as the language used in classroom differs from the exchange setting (e. g. easier vocabulary, slower tempo, non-native accent). Students therefore have difficulties when they are exposed to authentic language in an exchange setting and it takes time for them to adapt and to build social bonds. Face-to-face communication with native speakers is crucial to acquire a full repertoire of language skills, and, particularly, to be able to communicate successfully with native speakers in natural contexts. The author therefore advises to cyclically integrate exchange activities into the foreign language curriculum at school.

In Switzerland, there has only been one study on language exchange at primary level so far. The study by Botturi Kappler and Negrini (2018) examined the AlpConnectar project, an exchange project based on digital technology between nine primary schools (4th–6th grade) from the cantons of Grisons, Valais and Ticino. The exchange activities were spread out over a whole school year and included a virtual acquaintance phase on the one hand and the development of a more comprehensive didactic and interdisciplinary course unit on a topic in the field of biology, geography, and history by means of video sequences on the other hand. The aim was for learners to document certain aspects of their village or home in the foreign language and make them accessible to the other class. The project ended with an encounter of the respective classes involved. The interviews conducted with pupils and students indicated that a virtual language exchange in the Alpine arc between the German-, French-, and Italian-speaking parts of Switzerland had improved the language skills of the primary school pupils. In addition to improving their language skills, they expanded their communication strategies and reduced their inhibitions about communicating in the foreign language.

Apart from these studies at primary level, there are also a few studies at lower secondary school level. The study by Oakley et al. (2018) aimed at identifying challenges and successes associated with the design and implementation of a digital storytelling exchange between three Australian und four Chinese middle schools. 151 students (aged between 13 and 15) participated, 34 in Australia and 117 in China. Pupils in both countries were asked to work in small groups and to describe their daily lives and a traditional tale from their own culture in the target language (Mandarin for Australian students, English for Chinese students) in a multimodal format. The digital stories were posted on an online platform in video format, so that the overseas pears could respond, comment, ask questions or offer corrections or suggestions. Teachers were interviewed at the end of the project to elucidate their perspectives on what had worked well and what challenges they and their students had experienced. According to the teachers the digital stories enabled student learning through authenticity and real-world relevance. Tokaryk (2019) makes a similar observation in her eTwinning project. Pupils aged 12–15 years from 7 European countries, all beginners of French, wrote a story in French together using digital tools. According to the teachers, in doing so, the young people developed reading strategies, vocabulary, and discovered new grammar rules.

It must be noted that none of these studies assessed the language skills of the students involved by means of tests. The statements on linguistic competence and progress are based on self-assessments of the participants, teacher assessment, and in some cases also participant observations. Furthermore, the studies presented so far are not intervention studies with a control group. While they provide valuable information about the development of age-appropriate exchange settings, they do not allow for well-founded statements about the effectiveness of the exchange activities studied. This is particularly true for longer projects such as Alpconnectar, which extend over a whole school year, as it can be assumed that the language skills of the pupils would also have improved in regular language classes.

While Massler’s study (2008) is not based on a (quasi-)experimental design either, it is one of the few studies which examined the language competences using writing products of the participants. It showed that a project-oriented six-month e-mail contact between a class of German 9th grade pupils and a Canadian high school class had a positive effect on communicative and cooperative writing skills in English. Whether similar project-oriented writing lessons would have been just as effective without exchange with pupils from the target language area remains open, however, as no control group was examined. Evans and Fisher (2005) studied the effects of a six to eleven-day language stay of 68 English-speaking 13 to 14-year-old adolescents in France by analysing systematically pre- and post-departure tests on listening, reading, speaking, and writing. The greatest improvements were found in listening comprehension and writing (fluency and communicative content, but not accuracy). The study implied a comparison of the results taken two years later in the public GSCE examination with a matched control group. No long-term effects of the stay were found, but the analysis yielded a small correlation between the stay and listening comprehension. The study also showed a correlation between linguistic progress and the language support of the host family in terms of help and error correction, as reported by the pupils.

Noteworthy quasi-experimental studies on children that examined the language learning gains in exchanges settings using intervention and control groups include the study by Llanes (2012) and Llanes and Serrano (2017). Llanes (2012), which explored the short- and long-term effects of a two-month stay abroad (SA) on nine 11-year-old Catalan/Spanish bilingual’s oral and written skills, found significant gains in their oral skills immediately after the stay as compared to seven peers who had stayed at home (AH). Contrary to Evans and Fisher (2005), no significant increase in their written skills was found. Llanes and Serrano (2017) compared the effect of learning context (SA vs AH) on three different age groups: children (age 10–11), adolescents (age 12–15) and adults in terms of their oral and written development (N=197). All SA participants spent 2–3 months abroad. While the SA context did not affect the participants’ writing skills, it had an effect on their speaking skills with the SA context being more beneficial than the AH context. In terms of age, the adults showed greater increase in their written skills than the rest of the groups. In terms of oral skills, the differences were less pronounced. Here adults also outperformed the other groups but only with regard to lexical diversity. Regarding the interaction effects between learning context and age, younger participants in the SA context seemed to experience the greatest gains concerning the oral skills measured, but these effects were not so clear with respect to the written skills.

3 Methodology

3.4 Design and instruments

The productive language skills of the pupils were examined longitudinally using a quasi-experimental design with an intervention and control group and a pre-test post-test design. The data collection took place at two time points: in August 2017 before the start of the intervention (pre-test; t₀) and in May 2018 (post-test; t₁), approximately one month after the second encounter and a week after the last follow-up lesson. The same tasks were used at t₀ and t₁ and administered simultaneously to the intervention and the control group. For research economic reasons, the productive skills were only examined in the target language French (that is in the German-speaking intervention and control classes).

In addition to the language tests, a student questionnaire was administered before and after the intervention. It contained around 45 items and recorded socio-demographic data (gender, age, nationality, first languages), motivational aspects such as general language learning motivation, foreign language use anxiety and self-efficacy beliefs in relation to foreign language use. The development of these motivational aspects will not be discussed further below as this is done in another article (Heinzmann et al. 2022). Furthermore, a teacher questionnaire was administered to the teachers of both the intervention and control classes in order to gauge the teacher’s motivational dispositions and self-efficacy beliefs as well as the intervention class teachers’ perception of the effects of the intervention. Due to the focus on the competences of the students, this data source will not be discussed in detail in this article, but will be used to support the interpretation of the results.

3.4.1 Speaking tasks

The French speaking competences of the pupils in the intervention and control classes in German-speaking Switzerland were assessed by means of an interview and a picture comparison. The dialogic interview consisted of six simple questions for the pupils, such as Comment ça va?‘ How are you?’. The questions were based on the DELF Prim A 1.1 exam, whereby the vocabulary being used was adapted to the coursebook used. In the monologic picture comparison, the pupils had to name three differences in two similar but not quite identical pictures. The pictures used were taken from the Cambridge Young Learners Test (OCR 2006, OCR 2007; new Cambridge Assessment 2019). It was ensured that they could well be described by means of vocabulary that the pupils knew from their coursebook. The assessment of pupils’ speaking competences took between five and ten minutes per pupil. The interlocutors were researchers from the St. Gallen University of Teacher Education. Group assignment (intervention vs control group) was known to the testers. The pupils were given no preparation time.

3.4.2 Communicative writing task and C-test

The French writing competences of the pupils from the intervention and control classes were assessed with a communicative writing task from the lingualevel^[5] repository (unpublished lingualevel task). As part of this task, the pupils had to write a short e-mail with information about themselves. The contents that had to be addressed in the e-mail were specified in German. A C-test was used to assess the pupils’ general language competence. The C-test also originated from lingualevel and contained three short texts with 16 gaps each. The pupils had to fill the gaps with common French words (cf. Figure 1).

Figure 1

Excerpt from C-Test

As a further source of data, teacher questionnaires were used to collect information on the organisation of the exchange, the teachers’ qualifications and experiences as well as their attitudes and motivation with respect to carrying out the exchange. Furthermore, the researchers were present as participant observers at each physical encounter and took field notes.

4 Results

4.1 Speaking skills

The speaking test score was calculated by adding up the points achieved for the individual subtasks of the interview, the picture description and the points awarded for self-initiated tasks^[7] (cf. Table 3 and Table 4). The maximum achievable score was 36. The empirical test scores of the pupils ranged from 6 to 31 for t₀ and from 7 to 32 for t₁, which indicates that there is neither a ceiling nor a floor effect.

In both groups, there was a significant increase in the test score from the pre-test to the post-test. The mean values of the two groups at t₁ are practically identical, but the increase across time is significantly greater in the intervention group than in the control group which was confirmed by linear regression (cf. Table 7). It needs to be pointed out, however, that the control group started out with a significantly higher pre-test score than the intervention group.

Table 6

Speaking skills – Intervention and control groups t₀ and t₁ in comparison

Study group Mean value difference t 1 –t 0		N	Minimum	Maximum	Mean value	Standard deviation
Control group +4.20	Speaking skills t0	58	6	31	21.19	5.20
	Speaking skills t1	56	9	31	25.32	4.24
	Valid values	56
Intervention group +7.35	Speaking skills t0	107	6	28	17.79	5.93
	Speaking skills t1	103	7	32	25.14	4.51
	Valid values	103

Figure 2

Speaking skills – Mean values and confidence intervals t₀ and t₁

Table 7 presents a regression model with the difference in test scores between t₀ and t₁ as the dependent variable and the intervention and other significant predictors as independent variables. The intervention turns out to be a significant predictor for the development of pupils’ speaking skills. Additional explanatory variables are the speaking test score as well as learners’ language use anxiety at t₀, both of which have a negative impact on the development. The class mean of the speaking test score t₀ is also included in the regression to account for the clustered nature of the sample. However, this control variable is not significant with respect to the development of speaking skills. The model explains 59.7 per cent of the variance and thus indicates a large effect.^[8]

Table 7

Speaking skills – Regression models to explain the difference t₁–t₀

Model		B	Standard error	Standard Beta	T	Significance
R 2 = .597 F(4,148) = 55.813 p = .000	(Intercept)	17.72	1.92
	Intervention	1.75	.57	.19	3.08	.00
	Speaking skills t0	-.57	.05	-.71	-11.44	.00
	Speaking skills t0 (class mean)	.04	.09	.03	.47	.64
	Language use anxiety t0	-.99	.36	-.14	-2.72	.01

4.2 Writing skills

In the writing task, there were 7 subtasks that were rated according to the rating criteria in Table 5 and two additional points for following genre conventions^[9]. Therefore, the maximum score is 16 points. The students’ test scores for t₀ ranged from 0 to 14 and for t₁ from 3 to 16.

With respect to the writing skills, too, a significant increase in test scores could be observed between the pre-test and the post-test. The increase is comparable in both groups. However, once again the control group already had higher scores at the outset.

Table 8

Writing skills – Intervention and control groups t₀ and t₁ in comparison

Study group Mean value difference t 1 –t 0		N	Minimum	Maximum	Mean value	Standard deviation
Control group +3.48	Writing skills t0	58	1	14	9.31	3.11
	Writing skills t1	56	7	16	12.79	2.12
	valid values	55
Intervention group +3.87	Writing skills t0	107	0	14	7.82	3.22
	Writing skills t1	105	3	16	11.69	2.70
	valid values	103

Figure 3

Writing skills – Mean values and confidence intervals t₀ and t₁

Table 9 shows a regression model with covariates to predict the change in the writing skills test score.

Table 9

Writing skills – Regression models to explain the difference t₁–t₀

Model		B	Standard error	Standard Beta	T	Significance
R 2 = .669 F(4,142) = 72.898 p = .000	(Intercept)	3.60	1.17
	Intervention	.07	.31	.01	.23	.82
	Writing skills t0	-.75	.05	-.88	-16.57	.00
	Writing skills t0 (class mean)	.36	.11	.19	3.17	.00
	Language learning motivation t0	1.23	.21	.29	5.78	.00

In the model, which includes control variables, the intervention is not significant. Significant predictors that contribute to the explanation of the development of pupils’ writing skills are their writing skills at t₀, the class mean of writing skills at t₀ as well as their motivation to learn the target language at t₀. Once more, the test scores of the pre-test have a negative effect on development, i. e. learners who already have a high test score before the intervention show a lower increase on average. Additionally, class context plays a role in explaining the development of writing skills, as the class mean is a significant predictor. Furthermore, pupils with high motivation to learn the target language before the language exchange exhibit a more favourable development regarding writing skills. The corrected model explains 66.9 percent of the variance in the difference in test scores and is significant.

Lexical diversity: To determine lexical diversity, Guiraud’s Index for Lexical Complexity (RTTR) was calculated for the written texts of each pupil at both time points. In both groups, a significant increase in lexical diversity can be observed across time. The increase is somewhat greater in the intervention group than in the control group. However, here too, the control group already had a higher starting level at the outset.

Table 10

Lexical diversity – Intervention and control group t₀ and t₁ in comparison

Study group Mean value difference t 1 –t 0		N	Minimum	Maximum	Mean value	Standard deviation
Control group +0.49	RTTR t0	58	1.00	4.80	3.48	.64
	RTTR t1	56	2.12	5.98	3.97	.81
	Valid values	55
Intervention group +0.78	RTTR t0	107	1.41	5.37	3.15	.76
	RTTR t1	105	2.11	5.61	3.93	.76
	Valid values	103

In the regression model (cf. Table 11) the intervention turns out to be a significant predictor for the development of lexical diversity in writing. Besides that lexical diversity at t₀, the class mean of lexical diversity at t₀ and the motivation to learn the target language turn out to be significant predictors of development. The higher the language learning motivation at the outset, the more lexical diversity increases. The lexical diversity of the pre-test has a negative effect on development. Learners with high scores at the beginning show a lower increase on average. As was found for writing skills in general, the class context also plays a role in explaining the differences in development of lexical diversity. Students from classes with a higher class mean before the intervention increase their lexical diversity more. The model explains a total of 30.8 percent of the variance in the change in lexical diversity.

Figure 4

Lexical diversity – Mean values and confidence intervals t₀ and t₁

Table 11

Lexical diversity – Regression models to explain the difference t₁–t₀

Model		B	Standard error	Standard Beta	T	Significance
R 2 = .308 F(4,148) = 17.456 p = .000	(Intercept)	-24.40	5.82
	Intervention	3.74	1.17	.24	3.20	.00
	RTTR t0	-4.94	.90	-.45	-5.50	.00
	RTTR t0 (class mean)	10.09	1.78	.49	5.67	.00
	Language learning motivation t0	4.58	.81	.40	5.63	.00

C-Test: The test value was calculated as the sum of all correctly filled in gaps, so that a maximum score of 48 points could be achieved. Only gaps where each individual letter of the word filled in was correct and in the correct position were considered correct. The C-test was very difficult for the pupils, which is evidenced by the moderate test scores at both time points. These ranged between 0 and 29 for t₀ and between 2 and 34 for t₁.

Both in the intervention group and in the control group, a significant increase in the test scores across time can be observed. The increase is significantly higher in the intervention group than in the control group. It should be noted, however, that the test scores already differed between the two groups at the outset (t₀). The control group started with a significantly higher mean value before the intervention, which indicates a higher initial level in the control group.

Table 12

C-Test – Intervention and control group t₀ and t₁ in comparison

Study group Mean value difference t 1 –t 0		N	Minimum	Maximum	Mean value	Standard deviation
Control group +5.61	C-Test t0	58	0	29	16.53	5.26
	C-Test t1	56	9	33	22.14	5.58
	Valid values	55
Intervention group +8.62	C-Test t0	106	0	29	13.12	5.57
	C-Test t1	105	2	34	21.74	5.99
	Valid values	102

Figure 5

C-Test – Mean values and confidence intervals t₀ and t₁

In the regression model to predict the development in the C-test test scores (cf. Table 13), the intervention turns out to be a significant predictor despite the inclusion of control variables. In addition to the intervention, the pupils’ C-test scores at t₀, the class mean of C-test scores at t₀ as well as pupils’ self-efficacy beliefs and the foreign language use anxiety at t₀ proved to be significant predictors of the linguistic progress as measured by the C-test. Besides the C-test scores at the outset, the clearest influence on the development of overall proficiency are pupils’ self-efficacy beliefs at the outset: learners who believe in their foreign-language skills show a greater increase in their overall proficiency. The test result t₀ on the other hand has a negative effect on development, i. e. the higher the test score of a person at the outset, the more likely it is that their increase is lower. The class context determines part of the variance as the class mean at t₀ is significant. The language use anxiety, surprisingly, has a positive effect on the development of learners’ C-test scores: students with higher anxiety show a greater increase in test scores. Overall, the corrected model explains 29.1 percent of the variance and is significant.

Table 13

C-Test – Regression models to explain the difference t₁–t₀

Model		B	Standard error	Standard Beta	T	Significance
R 2 = .329 F(5,146) = 15.313 p = .000	(Intercept)	-7.92	3.36
	Intervention	2.96	.80	.33	3.68	.00
	C-Test t0	-.45	.07	-.52	-6.48	.00
	C-Test t0 (class mean)	.43	.16	.25	2.70	.01
	Self-efficacy beliefs t0	3.15	.57	.47	5.50	.00
	Language use anxiety t0	1.67	.51	.26	3.24	.00

5 Discussion

Action-oriented tasks were used to assess both speaking and writing skills. In addition, a C-test was used to measure general language competence. With regard to writing skills, results are mixed. While no effect of the intervention could be demonstrated for task fulfilment, a positive effect could be demonstrated for lexical diversity. As far as writing task fulfilment is concerned, significant progress was observed across time in both groups. This progress was comparable across groups, however, and did not differ significantly between the groups. This finding contrasts with the results of the Evans and Fisher (2005) study, where, however, the learners were slightly older and more advanced in their learning of French, and the stay was longer and more intensive. On the other hand, it is in line with the study by Llanes (2012) as well as Llanes and Serrano (2017), who also found no effect for the development of writing skills. However, the positive effects found for lexical diversity are in contrast to the study by Llanes (2012) as well as Llanes and Serrano (2017), where the same measure for lexical diversity was used and no effect of a stay abroad could be found. Given the mixed results of studies dealing with the effects of mobility programs on writing skills both with older and younger learners in a study abroad context (Pérez-Vidal and Barquin 2014; Pérez-Vidal and Juan-Garau 2009; Sasaki 2004, Sasaki 2007, Sasaki 2009, and Sasaki 2011), the mixed results in this study are not overly surprising. Llanes and Serrano (2017) point out that apparent contradictions could be related to the instruments used for measuring writing skills. As Evans and Fisher’s (2005) analyses suggest, qualitative analyses of written texts may capture more subtle morphosyntactic, lexical and idiomatic changes related to authentic communicative experiences in the target language. However, Llanes and Serrano (2017) also suggest that the inconclusive results might be attributable to the different length of programs investigated or the amount and type of practice that learners receive during the intervention. It may be that a longer stay or an intervention that is more clearly targeted at the promotion of writing skills would be more effective for their development. While the subjects in our intervention group also had written contact with the partner class throughout the school year, communication during the two physical exchange meetings was entirely oral. It can be assumed that pupils in regular French lessons (and hence the students in our control group) also repeatedly encountered writing occasions in their French classes. On the other hand, the pupils in the control group did not have the opportunity for repeated oral interaction with French-speaking pupils. This could explain why the intervention primarily had a positive effect on speaking skills.

Although with regard to speaking skills, too, significant progress can be observed in both groups across time, the progress is significantly greater in the intervention group (cf. Table 7). Consequently, the intervention is a significant predictor of the development of speaking skills. It needs to be pointed out, however, that the intervention and control group already differed significantly with regard to their speaking skills before the intervention. The pupils in the control group had a significantly higher starting level. However, the test values do not suggest a ceiling effect, i. e. the pupils were not able to solve all or almost all the test items. Accordingly, the test was not too easy for either group.

Apart from the positive effect on speaking skills and lexical richness in writing, the intervention also had a positive effect on pupils’ general language competence, which was assessed by means of the C-test. The picture here is analogous to that identified with respect to speaking. The test scores of both the control and the intervention group increased significantly from the first to the second test. However, the increase was significantly greater in the intervention group than in the control group (cf. Table 13). Once again, however, the starting level of the two groups differed significantly at the outset of the study (t₀) with the control group outperforming the intervention group, which indicates a higher initial level of language competence in the control group. Due to the relatively low test scores of both groups in the C-test even after the intervention, a ceiling effect can be ruled out.

The fact that learners with initially lower levels of competence made greater progress than learners with initially higher levels of competence reflects the results of two large-scale studies that looked at the predictors of learning gains in study abroad settings. In the study by Brecht et al. (1995) involving American students, those students who had lower test scores in speaking, listening and reading at the beginning of their stay abroad showed greater gains. The same pattern was found in a study by Lapkin et al. (1995), which focused on secondary school students. Again, those students with lower initial test scores in listening, speaking, reading and general language proficiency benefited more. However, Dewey (2007) notes that these findings could also be attributable to a ceiling effect and that the tests used were not demanding enough to detect progress by more advanced learners. Such a ceiling effect can be largely ruled out for the present study, however, based on the test scores achieved. On the other hand, the results of the present study on the influence of prior competences on learning progression also differ from the findings of two previous studies in which children and young people with a better starting position benefited more (Krok et al. 2010; Heinzmann et al. 2014). The question as to which learners benefit more or less from which exchange formats and why needs further clarification.

All in all, the present study shows a small positive effect of the exchange setting on the general language competence of the pupils, their speaking skills and their writing skill in terms of lexical richness. When carrying out a language exchange at primary level, it is certainly important not to overlook the fact that pupils are at an elementary level of language use only. Accordingly, an implementation at this level needs appropriate scaffolding measures on the part of the teachers and didactic materials as well as realistic expectations of progress on the part of pupils and teachers.

6 Limitations

Obviously, the results of the study have to be interpreted in light of certain limitations. Firstly, we were unable to create a randomized sample or to closely match control and intervention groups with regard to target and control variables due to the difficulty of recruiting classes. Therefore, a selection bias must be assumed to exist with teachers in the intervention group being those who were eager to participate with their class in an exchange activity. Also, the two groups already differed with regard to their language skills at the beginning (t₀), which might have had an influence on the results. A further methodological limitation is that both testers and raters were aware of the group assignment of the students that they tested/rated, which might have had an influence on their testing or rating behaviour. Thirdly, we used holistic rating criteria rather than fine-grained linguistic analyses for the evaluation of the pupils’ written and oral productions. Using the latter would have allowed for more detailed information on pupils’ progress in the foreign language. Furthermore, no literal transcriptions of pupils’ oral productions were made for research economic reasons for which reason it was not possible to investigate lexical diversity in students’ speaking skills in analogy with the analyses conducted for writing.

7 Conclusion

Despite its brevity and low-threshold nature, the intervention had a positive impact on the pupils’ speaking skills and general language competence as well as partly on their writing skills. This finding suggests that even short encounters at primary and beginner level can improve language competencies. As mentioned above, the design of the study does not allow any conclusions to be drawn about the extent to which the use of the accompanying material contributed to this result. Still, it can be stated that he majority of the teachers appreciated the authentic use of the foreign language and the fact that they worked towards the mastery of a specific encounter during the regular French lessons in which the acquired skills could be applied. They also appreciated the didactic material provided by the project team, which they found very useful, practical and helpful. Due to the high effort involved in preparing, implementing and following up a language exchange, it is important to familiarise student teachers in initial and in-service training with existing didactic materials and to make them available to other interested parties (see Krüger-Potratz et al. 2018)^[10]. In the interest of sustainability, it is also advisable to continue working with the class, teacher or school tandems after a successfully completed exchange activity, as this can also considerably reduce the effort required for a renewed implementation.

Another aspect, which needs to be considered in view of obtaining greater benefit from exchange projects, is the timing of such projects. The intervention took place in 6^th grade. The end of 6^th grade in Switzerland marks the transition from primary to lower secondary level and, thus, the dissolution of the existing school classes and rearrangement of classes into three different performance levels. As a result of the transfer, the tandem classes can no longer be brought together in the following year, so that the contacts once established between pupils are at best continued at a private level. Where this is possible, e. g. in smaller municipalities, it would make sense for the schools as institutions to continue the exchange activities initiated at primary level at lower secondary level. Given the superiority of children in benefitting from exchange and study abroad experiences (Llanes and Serrano 2017) and in the light of the encouraging results of Krüger-Potratz et al. (2018) in contexts with little previously developed target language competence, an earlier start of such activities could also be envisaged, provided sufficient scaffolding by accompanying teachers.

While our study points to the potential of short exchange activities to foster the development of productive skills of primary school children, it is clear that not every student profited to the same degree – as was also shown in other studies (e. g. Evans and Fisher 2005). This warrants future studies on the processes influencing why some students develop more than others in exchange settings. Finding out more about the processes behind learning and development also opens up paths for organizing and implementing exchange programs in such a way as to facilitate optimal learning conditions.