The ERASMUS sojourn: does the destination country or pre-departure proficiency impact oral proficiency gains?

ABSTRACT

This study examines the influence of pre-departure proficiency on English as a second language (L2) oral development after a semester studying abroad in two distinct learning contexts. The data analysed were collected from tertiary level students (N = 33) who undertook an ERASMUS exchange semester in a European country where English was either the official language (e.g. England) or used as a lingua franca in the university context (e.g. Germany). Each participant completed an oral elicited imitation test and a free oral response task pre-departure and on their return. The results of quantitative analyses indicate that the type of the study-abroad context had no significant impact on oral proficiency gains as all participants were found to perform significantly better on the elicited imitation test after the semester abroad. Despite strong to medium statistical correlations identified between pre-departure proficiency, and gains in oral fluency, lexical diversity, and syntactic complexity, pre-departure oral proficiency was only found to significantly influence syntactic complexity development as pre-intermediate and upper-intermediate participants outperformed intermediate students in both contexts.

KEYWORDS:

• Learning context
• pre-departure proficiency
• study abroad
• second language development
• complexity
• accuracy
• fluency
• Oral L2 development
• elicited imitation test

Introduction

The study-abroad (SA) context has long been considered advantageous for L2 learners due to the amount of input and interaction assumed to be available through in-class and out-of-class exposure to the target language (TL) (Borras and Llanes 2019). This assumption is hardly surprising given that the traditional SA context has been assumed to be a host country where the TL is the official language. However, particularly with the increasing number of exchange programmes happening between academic institutions across Europe, a slightly different context is emerging – one where the TL (English) functions as a lingua franca (Jenkins 2014) rather than an official language. I refer to this context as English-as-a-lingua-franca study abroad (ELFSA) (Köylü 2016).

Thus far, we have limited empirical research on ELFSA language learning and given its growth, it is certainly a context which merits further exploration (Llanes et al. 2016; Köylü 2016; Köylü and Tracy-Ventura, under review). Significant affordances of the ELFSA context could be, for example, the opportunity to learn the TL in an anxiety-free atmosphere where learners prioritise fluency and effective communication strategies, rather than obsessing over native speaker (NS) norms or the accuracy of their production (Köylü 2016). Further research is thus essential to understand the dynamics of L2 development in this learning context.

Thus, the current study aims to address this gap by comparing oral proficiency gains in a ‘traditional’ SA context where English is the official language, with those made by students in an ELFSA environment. It also explores the relationship between pre-departure proficiency and oral complexity, accuracy and fluency (CAF) gains from study abroad, irrespective of context.

Literature review

Comparing the two contexts

The literature comparing traditional SA with at-home (AH) formal instruction clearly indicates that SA positively affects L2 development, due to the high quality and quantity of exposure to the TL (see Borras and Llanes 2019 for a review). The SA has been investigated considering proficiency (Serrano et al. 2016), oral (McManus et al. 2020; Mora and Valls-Ferrer 2012) and written gains (Pérez-Vidal 2014), as well as a focus on several programme-related and socio-pragmatic factors, such as social networking (Mitchell et al. 2017). The ELFSA context, on the other hand, has thus far been investigated in relation to variables representing how tertiary-level learners interact with their setting such as language awareness (Martín-Rubió and Cots 2018), L2 beliefs (Kaypak and Ortaçtepe 2014), with some research on L2 oral (Llanes 2019; Martín-Rubió and Cots, 2018) and written development (Llanes et al. 2016). Contextual characteristics influencing learner identities of university-level ERASMUS exchange students (Kolacsai 2009) and beliefs about learning an L2 (Kaypak and Ortaçtepe 2014) have also been analysed. ELFSA was found to be beneficial in increasing linguistic awareness (Martín-Rubió and Cots 2018), changing learner focus from issues of accuracy to fluency (Kaypak and Ortaçtepe 2014), and in developing within an anxiety-free learning atmosphere, with students feeling comfortable interacting in the TL with like-minded non-native speakers (Borghetti and Beaven 2017). Llanes et al. (2016) looked into writing improvement in L2 English of 39 Catalan/Spanish bilinguals majoring in engineering after a semester of ELFSA and found that those with higher pre-departure proficiencies improved more. Similarly, Llanes (2019) investigated the oral gains of 18 Spanish-speaking ERASMUS university students after a semester abroad in an ELFSA context. The results indicated significant gains in oral fluency, lexical complexity, and proficiency as measured by the Oxford Placement Test (OPT).

Those few studies comparing ELFSA directly with traditional SA have found equal benefits for the former (Köylü 2016; Köylü and Tracy-Ventura, under review). In particular, Köylü and Tracy-Ventura (under review) compared oral and written CAF gains in English as an L2 after a semester in three different contexts: traditional SA, ELFSA, and an intensive AH programme. The tertiary-level sojourner participants as a whole significantly developed their oral speech rate, breakdown fluency, accuracy, and written fluency over time, and no differences were found between the SA and ELFSA groups. When compared, the AH group significantly outperformed the SA and ELFSA only on written fluency. Further research is now needed to differentiate linguistic gains made in the different SA contexts, especially in terms of some under-researched constructs, such as proficiency development and the effects of pre-departure proficiencies on L2 gains.

Oral CAF development in study-abroad research

Research into context and L2 oral development has typically yielded positive results for the traditional SA context, when compared with the AH context. The studies utilising CAF constructs to operationalise L2 oral development have found variously that traditional SA can be effective in developing oral fluency (Mora and Valls-Ferrer 2012); accuracy (Leonard and Shea 2017; McManus, Mitchell and Tracy-Ventura, 2020); lexical (Llanes 2019) and syntactic complexity (Leonard and Shea 2017). However, findings regarding CAF gains after the SA experience have not always been consistent, particularly for syntactic complexity and accuracy (Borras and Llanes 2019).

Looking at the learners’ beliefs regarding development in the ELFSA, Kaypak and Ortaçtepe (2014) found that sojourners came to prioritise fluency and effective communication. Similarly, investigating the case of university students with different majors, Köylü (2016) and Llanes (2019) noted that after a semester of study abroad within the ERASMUS scheme, their tertiary-level participants reported to value becoming fluent and effective communicators more than having error-free native-like speech. Focusing on L2 oral accuracy in the ELFSA context, Martín-Rubió and Cots (2018) found some gains for their participants majoring in education and engineering based on descriptive statistics, yet no statistical tests were utilised to check for any significant differences before and after the ELFSA experience. Llanes (2019) did check the statistical significance of accuracy gains before and after ELFSA but the reduction in errors per T-Unit that she found in her post-ELFSA participants’ oral performances were insignificant. These inconclusive findings concerning L2 oral gains in the ELFSA do suggest that further research is needed to check whether significant oral improvement takes place after a semester in an ELFSA context and how any improvement compares with improvement after a similar period in a traditional SA context. A further variable to be taken into consideration is the potential impact of pre-departure proficiency as this was found by Llanes et al. (2016) to influence L2 writing gains in an ELFSA context.

Operationalising and measuring proficiency

Proficiency has been investigated in SLA through a wide range of instruments, such as self-assessments, oral interviews, and standardised tests (Hulstjin 2011). In most cases, it serves as a secondary variable to be controlled when investigating another major construct (e.g. morphosyntactic or lexical development or effects learning context) (Tracy-Ventura et al. 2014). It has long been suggested that proficiency should always be included in SLA studies as it is a crucial intervening variable influencing the generalisability of results (Norris and Ortega 2012).

It is, however, a challenging task to define exactly what language proficiency entails and then attempt to measure it. It can be defined as ‘the ability to function effectively in the language in real-life contexts’ in its basic sense (Higgs 1984: 12 as cited in Leclerq et al. 2014: 6). However, with advances in SLA research, this definition has been revamped comprehensively. Hulstijn (2011) proposes a definition to encompass both L1 and L2 proficiency, making a distinction between two kinds of language abilities: basic language cognition (BLC, basic use of language for everyday communication) involving core language (e.g. higher-frequency words, and less complex morphosyntax); and higher language cognition (HLC, advanced uses of written and spoken language for e.g. academic purposes) with peripheral (e.g. complex morphosyntax) elements and functions (Leclerq et al. 2014). Thus, BLC as a language ability is the shared basis for competent daily oral communication among native speakers (NS) without much variation. However, native speakers show huge differences in their ability for HLC because they do not share the exact educational or occupational backgrounds, or personal interests (Tracy-Ventura et al. 2014: 144).

When it comes to L2 speakers, Hulstijn (2011) suggests that HLC can be acquired through education and training, but the acquisition of BLC after puberty is open to debate (Hulstijn 2011). Tracy-Ventura et al. (2014) underscore that when SLA researchers opt for measuring L2 proficiency, they consider BLC as their major construct and tend to measure it via ‘assessments of oral language using tests that are literacy-independent and that focus on operations involving high-frequency during the integrative use of oral language, while recruiting in real time all interrelated basic linguistic levels of lexicon, grammar, phonotactics and prosody’ (Tracy-Ventura et al. 2014: 144). The current study operationalises proficiency as a form of BLC and employs an elicited imitation test (EIT) (Ortega et al. 1999) on the grounds that it is a literacy-independent measure of BLC to document implicitly-acquired language competencies (Tracy-Ventura et al. 2014).

Elicited imitation test

The EIT is a test of sentence repetition used to tap global oral proficiency (Ortega et al, 1999). The test-takers listen to a TL sentence and repeats as much as they can. The stimulus sentences include more syllables as the test proceeds. The test draws on the theory that speakers can only accurately repeat what they have heard if they understand and can parse the target sentences through their developing grammars (Tracy-Ventura et al. 2014). Tracy-Ventura et al. (2014) touch upon issues of criticism regarding EITs as measuring the capacity of short-term memory span through a form of ‘parroting’ by imitating the sentences heard rather than a measure of language capacity through comprehension. However, these tests have been confirmed to be reliable and valid proficiency measures through empirical studies (Kim et al. 2016).

SA, pre-departure proficiency, and proficiency development

Research into SA reflects the tendency to conceptualise general proficiency as BLC, with widespread use of the Oral Proficiency Interview (OPI) as a measure of proficiency (Segalowitz and Freed 2004; Tavakoli 2018). Various literacy-independent productive tasks, such as picture-narration (Llanes and Muñoz 2009), monologues and dialogues (Tavakoli 2018) have also been used to measure development in proficiency. Given the never-ending discussion around how to measure proficiency and the affordances provided by SA contexts for providing rapid development opportunities for TL oral competence, further research using different instruments has long been called for (Hulstijn 2011).

There is particular interest in exploring issues considering measuring L2 proficiency specifically pertained to the ERASMUS scheme. An applicant to an undergraduate ERASMUS exchange programme offered in English has to document at least a pre-intermediate level of proficiency (B1, according to Common European Framework of Reference for Languages, CEFR) in the TL, and sometimes the host institution may require a higher level (e.g. UK institutes require at least intermediate B2 level). This has stimulated scholarly interest in measuring differences before and after the sojourn period as well as exploring the relationship between pre-departure proficiency and L2 development.

A few studies have chosen to measure pre-departure and post-sojourn proficiency using standardised literacy-dependent tests providing a single holistic score (Llanes et al. 2016; Llanes 2019), such as the Oxford Placement Test (OPT) or TOEFL (Tanaka and Ellis 2003). The majority, however, have employed literacy-independent task-based assessment, the results of which have typically indicated significant gains after study abroad (Segalowitz and Freed 2004; Tavakoli, 2018). EITs, on the other hand, have only recently been used in SA studies by Serrano et al. (2016). These researchers measured sentence repetition as a form of language performance, rather than as a measure of proficiency, using a shorter version of the EIT, comprising 20 sentences totalling 119 words. They compared teenagers learning L2 English after intensive short-term programmes of 3–4 weeks in AH and SA settings, and found no significant differences in EIT scores before and after the treatment. However, they did find that the SA context was more beneficial for oral lexical richness gains, while AH was slightly more advantageous for grammar (Serrano et al. 2016).

It is equally important to measure pre-departure proficiency in order to document participants’ actual threshold levels so as to be able to investigate any post-sojourn L2 gains. With this aim, scholars have explored the case of sojourners with both lower and higher levels of pre-departure competence (Llanes 2019). The idea that a certain level of TL knowledge is essential in order to benefit the most from exposure to L2 input in the SA context compared to little to no pre-departure knowledge is widely accepted in the literature (DeKeyser 2007; Leonard and Shea 2017; Llanes and Muñoz 2009; Valls-Ferrer and Mora 2014). DeKeyser (2007) theorises this by saying that without declarative knowledge before a period in a TL context, the learner will have nothing to proceduralise despite the high amount of input and output, but will make use of this exposure to build declarative knowledge.

Considering what level of pre-departure proficiency would be optimal, SLA research has yet to yield consistent findings. Some studies have shown that lower-level learners benefited more from a SA experience than higher levels (Llanes and Muñoz 2009; Valls-Ferrer and Mora 2014) while several studies have indicated the opposite - higher levels benefiting the most from the SA (DeKeyser 2010; Leonard and Shea 2017) - or no relationship; Llanes (2019) found no correlation between pre-departure proficiency levels and the amount of L2 gains over a semester in an ELFSA context. It has been pointed that results in favour of lower proficiency levels might depend on the selection of research instruments, in this case the OPT (Llanes 2019). Investigating a group of engineering majors, Llanes et al. (2016) explored whether pre-departure proficiency resulted in different written gains in the ELFSA, finding significant results for those with higher pre-departure proficiencies who ended up with more gains in grammatical complexity.

In light of these empirical findings, the current study aims to contribute to the SA literature by exploring the influence of context (traditional SA vs. ELFSA) on L2 proficiency, and specifically oral development after a semester abroad. Taking a comparative perspective into L2 development, this study is also of significance in that it employs a large collection of measures investigating all aspects of oral L2 performance as speech rate, breakdown and repair fluency, syntactic and lexical complexity, and accuracy. It also seeks to gain insights into relationships between pre-departure proficiency and L2 oral development during the SA period. The following research questions direct the study:

1. To what extent does the SA context (traditional SA vs. ELFSA) have an effect on proficiency development over time?

2. Is there a relationship between pre-departure proficiency and oral gains over time?

3. If yes, to what extent does the level of pre-departure proficiency have an effect on oral performance of English as measured by complexity, fluency, accuracy, and syntactic and lexical complexity (CAF) development over time?

Methodology

Design

The results presented in this paper come from a subset of participants and data analysed in a larger project comparing SA and ELFSA with AH (Köylü 2016). The present study focuses on the influence of two independent variables, sojourn context and pre-departure proficiency (measured at Time 1 – prior to studying abroad) on two sets of dependent variables, oral CAF gains (formulated as the difference between Time 1 - prior to studying abroad and Time 2 – upon completion of the semester abroad) and post-sojourn proficiency (measured at Time 2 – upon completion of the semester abroad). The relationship between pre-departure proficiency and CAF scores was also investigated through correlation analyses.

Contexts and participants

A total of 33 (24 female, 9 male) sojourners with no previous SA experience took part in the study. All were L1 Turkish speakers learning English as an L2 studying a variety of majors, such as engineering, humanities, and social sciences in a high-ranked Turkish university. Their English proficiency was assessed at CEFR B1 level (intermediate) as part of the ERASMUS requirements, based on institutional tests at their home universities before departure. The age range was 20–30 years (M = 22). The traditional SA group had 9 participants who undertook their ERASMUS exchange semester (16-weeks) in England. The ELFSA group comprised 24 participants (numbers in Austria = 2, Czech Republic = 1, Denmark = 1, Finland = 2, Germany = 4, Greece = 1, Holland = 5, Italy = 5, Poland = 2, Portugal = 1). It was not possible to set up equal numbers of participants in the two context groups. Turkish universities have fewer mutual learning agreements with universities in the UK. Additionally, the high cost of living in the UK results in more Turkish students to prefer an ELFSA country (Turkish National Agency, 2011). To overcome issues of unequal sample sizes, robust statistical techniques, such as bootstrapping, was used to ensure reliability of results.

As part of their exchange programme, all the participants had to attend 9–12 h of classes per week at their host universities. However, they reported different degrees of workload in their host institutions, some reporting studying hard, some defining the experience as an ‘academic holiday’ (Köylü, 2016). Due to visa conditions, they were not allowed to engage in paid employment.

Tests

This study used two tests to obtain L2 data from participants: (1) an elicited imitation test (EIT, Ortega et al., 1999) to measure proficiency and (2) an oral performance task to assess development of L2 oral production. The tests and measures are explained below. All tests are publicly available via IRIS digital repository (https://www.iris-database.org/iris/app/home/index).

Elicited imitation test (EIT)

Although the participants had taken a pre-departure proficiency test at their home institutions, the researcher collected EIT data to explore their threshold competencies. These institutional tests largely vary in terms of difficulty and item types (multiple-choice vs. open-response). Thus, EIT was used to get a standardised pre-departure score for all the participants. The version of the EIT in English used was developed by Ortega et al. (1999). It includes 30 aural stimuli sentences increasing in number of syllables (see Appendix 1 for the list of EIT items). The researcher adapted the test for Turkish L1 participants with the inclusion of instructions and model sentences in Turkish to help them understand the test procedure. In between each sentence, a two-second pause was inserted followed by a half-second cue prompting repetition. The final version of the EIT utilised took 9 min 15 s including instructions and the practice section of 104 s.

The theoretical rationale for using the EIT test as a measure of proficiency is that a test taker can only repeat a target stimulus accurately if they have acquired some competence in the structures in the stimulus sentence (Ortega et al., 1999). The practical reason for selecting such a proficiency test was that it is a valid, reliable, short and practical test confirmed via empirical studies (Tracy-Ventura et al. 2014). To score the output, the original rubric developed by Ortega et al. (1999) was used (see Appendix 2). This allows a maximum score of 120 (30 sentences x 4 points each).

To ensure reliability, two near-native raters coded and scored the data. After they had managed to resolve any disagreements, a high statistical reliability was established for items (Cronbach’s α for traditional SA group = .964; Cronbach’s α for ELFSA = .969) and scoring (Cronbach’s α for traditional SA = .980; Cronbach’s α for ELFSA = .976). The EIT was administered twice: once pre-departure (pre-test) and once post-sojourn (post-test).

Free oral response task

Oral performance data were elicited by means of a one-minute TOEFL speaking prompt: What would you like to do in your free time and why? This test was administered online through videoconferencing twice, once pre-departure and once post-sojourn. It was felt that this topic would be easy to talk about, even at lower levels of competence. A short biographic interview took place as a warm-up activity before this task. All participants were given a 10-second preparation time after the prompt was revealed.

Procedures

The international offices of Turkish home universities responsible for mobility and exchange programmes were contacted to recruit participants, who were selected among volunteers with no previous sojourn experience. Before departure, all participants were contacted in order to arrange for pre-tests to be administered. After a 16–18-week interval, on return from study abroad, participants were contacted again for the administration of the post-tests. All data collection was completed online through videoconferencing.

Measures

The spoken data from the oral response task were first transcribed, then transformed into CHAT format and pre-analysed for CAF measures via CLAN (MacWhinney 2000) to be entered into statistical tests as dependent variables.

Complexity

Complexity was operationalised as syntactic complexity and lexical diversity:

• Syntactic complexity - (CL/ASU). This was determined through the ratio of finite and infinite clauses per Analysis of Speech Unit (ASU).

• Lexical diversity – (D). This was determined through the D values obtained by the VocD programme in CLAN (MacWhinney 2000) for each participant. This measure derives from a mathematical formula used to determine the frequency of the vocabulary used in the sample (low or high). Taking random samples from the text, D is calculated by taking type-token ratios of these short excerpts which are later used to create a random-sampling type-token ratio for the text. Scores range from 0 to 100 with higher scores indicating a greater diversity of vocabulary.

Accuracy – (ERR/ASU)

Accuracy was determined by counting the total number of grammatical and lexical errors per ASU. The data were segmented and coded by two researchers fluent in both Turkish and English to ensure reliability. All disagreements were reviewed and resolved between the two coders. A high inter-coder reliability percentage (92%) was reached for all measures.

Fluency

Oral fluency measures were operationalised, following Skehan (2009), as those dimensions of utterance fluency subject to the temporal aspects of speech. The three subdimensions measured were as follows:

• Speed fluency (W/T – words/time in seconds). The total number of words excluding words in disfluent production, divided by total production time in seconds, also referred to as ‘pruned speech rate’.

• Breakdown fluency (P/T – pauses/time in milliseconds). The total duration of silent and filled pauses longer than .250 milliseconds, divided by the total time expressed in seconds.

• Repair fluency (D/T – disfluencies/time in seconds x 60). The total number of disfluencies as determined by the number of repetitions, retraces, reformulations divided by total time expressed in seconds and multiplied by 60.

Statistical analyses

The independent, or the between-subjects, variables in this study were the context groups (traditional SA vs. ELFSA) and participants’ level of pre-departure proficiency (EIT scores). The dependent variables were pre-test vs. post-test EIT scores and pre-test vs. post-test CAF scores from the free oral response task, calculated through CLAN. Time was the within-subjects variable.

To discern inter- and intra-group development over time, a series of statistical tests were computed via the Statistical Package for Social Sciences (SPSS) version 27. Given the robustness of ANOVA and the fact that the same tasks were undertaken at two different times, a series of two-way mixed between-within repeated-measures ANOVA was used to see if there were any significant differences before and after the study abroad period in the two groups of participants (Field 2013). Before computing RM-ANOVAS, the dataset was tested for violations of analysis of variance, such as normality of distribution, equality of variances, and data sphericity.

The results of these preliminary analysis indicated that the data sphericity criterion was partially violated; thus, the values with Greenhouse-Geiser correction were reported to provide robust statistics (Field 2013). Additionally, the data for pre-test accuracy (ERR/ASU) and pre-test breakdown fluency (P/T) were found to be non-parametric. Hence, log-transformations were utilised to follow up with parametric tests on these variables (Field 2013). Further, to increase robustness given the unequal sample sizes of the context groups, all tests were run with bootstrapping with confidence intervals.

For the correlation analyses, the dataset was tested for violations to assumptions of correlation, such as normality of distribution and linearity. The data were found to be parametric with insignificant Kolmogorov–Smirnov results and a linear relationship was observed among the variables, having examined multiple scatterplots (Field 2013).

Results

RQ1: The effect of study abroad context on general proficiency development over time

Table 1 presents the descriptive statistics for the results of the pre-sojourn and post-sojourn EIT scores from the two context groups. We note that the traditional SA participants’ performance on the pre-departure EIT was lower than that of the ELFSA group. Comparison of pre- and post-sojourn means for both contexts shows that both groups increased their scores during the SA period.

Table 1. Descriptive statistics for EIT scores.

	SA (n = 9)		ELFSA (n = 24)
	M	SD	M	SD
Pre-EIT	70.33	24.49	81.29	24.81
Post-EIT	85.11	5.86	93.50	3.59

To see if this increase was statistically significant, a RM ANOVA was computed with context as the between-subjects variable, and time (pre- and post-) as the within-subjects variable. The results of this test yielded a significant main effect of time (F(1,31) = 34.241, p = .000, partial η² = .525), but not of context (F(1,31) = 1.438, p = .239, partial η² = .044). The interaction between time and context was also insignificant (F(1,31) = 21.607, p = .581, partial η² = .010). These results demonstrate that all participants significantly developed their proficiency over time regardless of context with a very large effect size (partial η² = .525), showing how influential SA time is on L2 development.

RQ2: The relationship between pre-sojourn proficiency and post-sojourn gains in oral CAF

The relationship between participants’ pre-sojourn proficiency (as measured by the pre-departure EIT) and their CAF gains on the oral response task after a semester abroad was statistically explored through correlation analyses using bootstrapping with CIs. Table 2 presents the results while Table 3 provides the descriptive statistics for the CAF scores.

Table 2. Correlations for pre-departure EIT and post-CAF scores.

Measures	R	p
Fluency (Speech rate)* Pre-EIT + W/T	.550	.001
Fluency (Breakdown fluency) Pre-EIT + P/T	−0.33	.007
Fluency (Repair fluency) Pre-EIT + D/T	−0.25	.180
Accuracy Pre-EIT + ERR/ASU	−0.2	.289
Lexical diversity* Pre-EIT + VocD	.547	.001
Syntactic Complexity* Pre-EIT + CL/ASU	.444	.012

*Significant correlations

Table 3. Descriptive statistics for oral production data by proficiency group.

Oral production data		PRE (n = 13)		INT (n = 10)		UP (n = 10)
		M	SD	M	SD	M	SD
Speech rate (W/T)	Pre-test	1.19	.337	1.64	.458	1.82	.341
	Post-test	1.55	.365	1.96	.456	2.10	.587
Breakdown fluency (P/T)^a	Pre-test	.428*	.188	.267	.121	.266	.090
	Post-test	.269*	.083	.207	.103	.205	.085
Repair fluency (D/T)	Pre-test	3.29	2.36	4.13	3.25	5.05	2.98
	Post-test	5.13	2.97	4.11	3.72	4.02	2.35
Accuracy (ERR/ASU)^b	Pre-test	.597	.332	.698*	.668	.251	.241
	Post-test	.334	.358	.257*	.124	.172	.175
Syntactic complexity (CL/ASU)	Pre-test	1.83	.639	2.70	.590	2.39*	.591
	Post-test	2.04	.801	2.20	.547	2.97*	.080
Lexical complexity (D)	Pre-test	29.86	16.7	40.14	10.9	40.56*	9.11
	Post-test	35.20	9.65	39.99	11.45	50.17*	11.49

^a A decrease in this score indicates fewer pauses.

^b A decrease in this score indicates more accurate production.

*The largest gains in comparison to other groups.

Significant large positive correlations with large to medium effect sizes were found between pre-EIT and gains in speech rate (W/T) (95% CI: 1.67, 2.02; r = .547, p = 000, R² = .299, d = 1.31) and lexical diversity (D) (95% CI: 36.8, 45.2; r = .547, p = 000, R² = .300, d = 1.32), while a medium positive relationship with a large effect size was found between pre-EIT and syntactic complexity (CL/ASU) gains (95% CI: .18, .36; r = .444, p = 000, R² = .197, d = .99), explaining 79.6% of total variance in the data.

RQ3: The effect of the pre-departure proficiency on CAF development over time

The strong positive relationships identified above led the researcher to investigate the data further for the effects of pre-departure proficiency on CAF gains after a semester abroad. As the sample size was small for a robust regression analysis and no significant differences were found for pre-departure proficiency between the two groups via independent samples t-test (t(1,32 = −1.134, p = .266), the sample was regrouped according to their pre-departure EIT scores. Four proficiency groups were determined based on intervals of 30 points: given a total of 120 possible points on the EIT, 0–29 was classed as elementary, 30–59 as pre-intermediate [PRE], 60–89 as intermediate [INT], and 90–120 as upper-intermediate [UP]). No participants, however, had scores in the elementary classification, so three groups – PRE, INT and UP – were used. To see the differences in CAF scores over time for each proficiency group, a series of RM ANOVAs were computed per pre-test - post-test pairs.

The results indicated that there was a significant interaction effect of time*pre-departure proficiency on syntactic complexity (F(2, 28) = 3.421, p = .012, partial η² = .272) with a large effect size. A series of Games-Howell post hoc comparison tests were computed to see where these differences lay. The results indicated that on the measure of syntactic complexity, the UP group (M change = .6558, p. =.000) showed greater gains than the PRE group (M change = .6558 vs .2908, p = .019), who in turn showed significantly greater gains than the INT group (M change = -.5062, p = .046). In other words, those with an upper-intermediate level of pre-departure proficiency showed significantly greater gains post-sojourn in subordination in their free oral production. Those with pre-intermediate levels of pre-departure proficiency also ended up demonstrating syntactic complexity gains, indeed greater than those who started their sojourn at intermediate level.

Only significant main effects of time were found for accuracy (F(2, 28) = 9.512, p = .005, partial η2 = .254), speech rate (F(2, 28) = 24.426, p = .000, partial η2 = .466), and breakdown fluency (F(2, 28) = 14.170, p = .001, partial η2 = .336) with large effect sizes. These results indicated that only time was significant on these constructs regardless of pre-departure proficiencies.

Discussion

Proficiency development

The current study explored the influence of (1) SA context and (2) pre-departure proficiency on sojourners’ oral development, measured by a general proficiency test (EIT) and complexity, accuracy and fluency (CAF) scores from a free oral response task. Providing one of the first empirical studies comparing the two different study abroad contexts (see also Köylü and Tracy-Ventura, under review), the results are of significance especially for those who are considering which destination country to choose for study abroad as learners of English. Firstly, the findings indicate that regardless of the context (traditional SA in an official Anglophone environment vs. ELFSA where English is used as a lingua franca), time spent abroad has a positive effect on proficiency. This contributes to the broader SA literature pertaining to proficiency (Segalowitz and Freed 2004; Llanes 2019). However, these significant results should be taken into consideration with a grain of salt due to the small sample size in the traditional SA group and its homogeneous nature although all possible statistical measures were taken to ensure robust results in the statistical tests.

We need to note that ERASMUS students, be it in the traditional SA or the ELFSA context, are engaged in formal learning in the TL (English) during their study abroad period. They attend university-level classes for at least 9 h a week, and their studies count towards their degrees at their home universities. Such an involvement with advanced academic work in the TL may explain why we see clear evidence of an improvement in L2 general proficiency and oral production in both the traditional SA and the ELFSA contexts. DeKeyser’s (2007) Skill Acquisition Theory (SAT, DeKeyser 2007) may help explain this, positing that large amounts of input and meaningful interaction in the L2 help learners proceduralise their declarative knowledge irrespective of the context of exposure, i.e. the status of the L2 in the host country.

The relationship between pre-departure proficiency and CAF gains

This study also investigated the relationship between pre-departure general proficiency as measured by EIT and oral L2 CAF development on return. The results of the correlation analyses indicated significant positive relationships between the proficiency measure (pre-EIT scores) and post-sojourn oral fluency in the form of speech rate, lexical diversity and syntactic complexity in the form of subordination. This suggests that the level of L2 proficiency on departure may influence the degree of development that can typically be achieved through a semester abroad.

These results, however, only allow us to confirm that there is a strong relationship between pre-departure proficiency measured by EIT and performance gains in speech rate, lexical diversity, and syntactic complexity after a semester abroad. The exact nature of that relationship requires further investigation. Given the nature of EITs (Ortega et al. 1999), a link between oral repetition capacity and oral fluency gains makes sense intuitively. Then the link between EIT and lexical and syntactic complexity may be explained via the importance of working memory, one’s developed interlanguage, and processing speed; in other words, how well a test-taker can store the oral stimuli while processing new information considering her linguistic knowledge (Kim et al. 2016). These findings echo Kim et al.’s (2016) research which found significant correlations between EIT and fluency and syntactic complexity.

The effects of pre-departure proficiency on CAF gains

The results indicated significant interaction effects for syntactic complexity, suggesting that different proficiency groups behaved differently over time resulting in differentiated gains. Generally in line with the literature (DeKeyser 2010; Mora and Valls-Ferrer 2012; Serrano et al. 2016) but contrary to the findings of Llanes and Muñoz (2009), the upper-intermediate group outperformed lower-level learners showing greater subordination gains over time. Interestingly, the pre-intermediate group also significantly outperformed the intermediate group unlike the findings in favour of lower levels (Llanes and Muñoz, 2009). It may be that intermediate level learners, having a mediating level of competence, are able to focus on aspects of L2 proficiency other than subordination or morphosyntactic complexity, such as fluency or accuracy and thus develop less in these areas than those with a slightly higher and a slighter lower level of proficiency.

Despite not detecting any interaction effects between time and pre-departure proficiency, time was found to significantly affect gains in accuracy, speech rate and breakdown fluency. This result should be interpreted carefully, recognising that the lowest level of proficiency of sojourners in this research was pre-intermediate; the results might be different with elementary-level learners. Thus, we can only conclude from our findings that for learners of pre-intermediate, intermediate, and upper-intermediate levels of proficiency, a semester-long SA (whether in a traditional SA or an ELFSA context) can be beneficial for promoting less error-prone, more fluent speech with fewer filled or unfilled pauses. Other individual factors, such as motivation, age, or personality might be more influential on gains after the sojourn experience.

In the case of ELFSA students, the results indicating a significant relationship between time and pre-departure proficiency contrast with findings from other ELFSA studies on oral (Llanes 2019) and written gains (Llanes et al. 2016). This might be explained by the fact that this study dealt with participants at three different proficiency levels whereas both Llanes (2019) and Llanes et al. (2016) looked at participants with a similar level of proficiency. Further, the test used to establish proficiency was different in these studies (OPT, giving a standardised holistic score, vs. EIT). Overall, in this study, the SA experience was beneficial for participants at all levels, with an additional advantage for syntactic complexity development for pre- and upper-intermediate levels. Further research is required to investigate the link between proficiency level, context, and L2 gains over time.

Conclusion

The current study investigated the effects of two study abroad contexts for L2 English learners – conventional SA in a host country with English as the official language, and ELFSA in a host country where English was used as a lingua franca in the university context. It also investigated the possible impact of pre-departure proficiency on oral L2 CAF development over time and the relationship between these variables. The findings contribute to the SLA literature in several ways. Time spent in the sojourn context, regardless of whether it was traditional SA or ELFSA, was found to positively impact proficiency. Thus, both contexts would appear to be valuable for L2 development – a finding that may help students preparing for study abroad to feel confident about choosing an ELFSA context.

Considering several limitations of the current study, such as sample size, tasks and measures utilised, further empirical insights are still needed. SLA researchers should investigate participants from different proficiency levels in comparative designs using a variety of monologic and dialogic tasks to tap oral and written development after SA, ensuring a robust sample size in all groups. Further research is still needed to see if pre-departure proficiency can predict L2 gains after the sojourn experience through advanced mixed modelling and regression analyses, which were not possible for this study due to the small sample size.

Furthermore, some qualitative measures should be incorporated in future research to see how sojourners interact with their environment so as to end up with gains. Also, the ELFSA countries included in this study may not reflect a ‘standard’ ELF exposure, varying from a more native-like variety in Scandinavia to a more contact language type in the Mediterranean. Thus, researchers might include a more equally dispersed country selection when investigating ERASMUS students.

This study also has several pedagogical implications. With improvements for accuracy, speech and breakdown fluency seen at all levels after study abroad in both contexts, along with additional subordination gains for pre- and upper-intermediate levels, this study calls into question the view that only people with higher pre-departure proficiency can benefit from a sojourn experience. Participants with lower competencies on departure completed the sojourn with a richer lexicon, more complex morphosyntax, and fewer instances of disfluency in speech; in addition, accuracy gains were significantly higher for intermediate sojourners. Thus, students with lower levels should not be deterred from taking part in an exchange programme. By the same token, more advanced students may need to be aware that they may not make equal progress in all areas of L2 language development; in this study, intermediate students did not progress in the area of syntactic complexity to the same extent as upper-intermediate and pre-intermediate students. Nevertheless, a study-abroad period, be it in a traditional SA or ELFSA, is still likely to be a valuable experience with both linguistic and non-linguistic benefits.

Disclosure statement

No potential conflict of interest was reported by the author(s).