Oral Interaction in the EFL Classroom: The Effects of Instructional Focus and Task Type on Learner Affect

Little is known about the effects of different instructional approaches on learner affect in oral interac- tion in the foreign language classroom. In a randomized experiment with Dutch pre-vocational learners ( N = 147), we evaluated the effects of 3 newly developed instructional programs for English as a foreign language (EFL). These programs differed in instructional focus (form-focused vs. interaction strategies- oriented) and type of task (pre-scripted language tasks vs. information gap tasks). Multilevel analyses revealed that learners’ enjoyment of EFL oral interaction was not affected by instruction, that willingness to communicate (WTC) decreased over time, and that self-confidence was positively affected by combining information gap tasks with interactional strategies instruction. In addition, regression analyses revealed that development in learners’ WTC and enjoyment did not have predictive value for achievement in EFL oral interaction, but that development in self-confidence did explain achievement in EFL oral interaction in trained interactional contexts.

ability to use this language in specific contexts (e.g., Celce-Murcia, 2007). Actual engagement in communication, however, is not only dependent on speakers' ability, but also on their affective state. The most well-known affective variables related to communication are willingness to communicate and self-confidence (e.g., MacIntyre, 2002). Willingness to communicate (WTC) is defined as the probability of engaging in communication when free to do so (McCroskey & Baer, 1985). While WTC is considered a stable personality trait in the first language (L1), Mac-Intyre et al. (1998) argue that one's willingness to communicate in a second language is likely to vary across situations, for example depending on one's familiarity with the interlocutor, and the formality of the situation. In a second or foreign language context, WTC can therefore be defined as the intention to engage in discourse with specific people in specific situations. MacIntyre et al. (1998) posit that this intention is primarily determined by the self-confidence one feels that communication will be successful in a specific situation, and conceptualize self-confidence as a combination of speakers' perceived communicative competence on the one hand and low levels of anxiety in using the second language on the other (Clément, Gardner, & Smythe, 1977). The link between selfconfidence and WTC has been established in several studies (e.g., MacIntyre & Charos, 1996;Mac-Intyre, MacMaster, & Baker, 2001;Yashima, 2002).
The adverse effect of anxiety on foreign language (FL) communication has been widely reported in literature (for an overview, see Horwitz, 2010), but enjoyment, which is seen as a positive emotion that runs parallel to the negative emotion of anxiety (Dewaele & MacIntyre, 2014) has only recently received attention in research (Dewaele & MacIntyre, 2014;Dewaele et al., 2018;MacIntyre & Vincze, 2017). High levels of foreign language enjoyment are linked to low levels of foreign language anxiety overall (Dewaele et al., 2018;MacIntyre & Vincze, 2017), and correlate positively with learners' self-confidence (Mac-Intyre & Vincze, 2017). Although this relatively new body of research does not focus specifically on the role of enjoyment in FL interaction, it shows that enjoyment is related to self-confidence, which plays an important role in FL interaction. MacIntyre (2002) argues that emotional arousal determines whether learners are energized into taking action or not. Positive emotions are more strongly implicated in this process than negative emotions (MacIntyre & Vincze, 2017). Thus, following Ajzen & Fishbein's (1980) Theory of Reasoned Action on which the notion of WTC is based, learners who feel confident and who enjoy interacting with specific people in specific situations are more likely to engage willingly in interactional activities than learners who feel anxious or who do not gain a sense of enjoyment from such interactions (cf. Kang, 2005).
Because language ability does not ensure language use, MacIntyre et al. (2003) posit that a fundamental goal of L2 instruction should be to produce learners who are willing to use the L2 in authentic communication, and good effects to this end have been achieved as a result of immersion programs (MacIntyre et al., , 2003. To date, however, little is known about the effects of different types of non-immersion FL oral interaction pedagogies on learner affect in FL interaction. This study thus seeks to establish how WTC, self-confidence, and enjoyment of FL oral interaction can best be fostered in the regular language classroom. Since research so far has not revealed clearly defined dependencies among these factors, we consider these variables as three related, but distinct factors in the present study. An additional question that arises is whether development in WTC, self-confidence, and enjoyment for oral interaction explains eventual achievement in EFL oral interaction. Of the three affective variables central to this study, the link between self-confidence and oral performance has received most attention in research. Selfconfidence is known to correlate both with the quantity of speech production (Dörnyei & Kormos, 2000;Phillips, 1992) and quality (MacIntyre, Noels, & Clément, 1997), and with use of compensation strategies (Liu, 2013;Yang, 1999), whereas lack of confidence in oral performance affects speakers' attempts to convey less concrete messages (Steinberg & Horwitz, 1986).
Studies into the role that enjoyment plays in oral ability are scarce. MacIntyre and Vincze (2017) report positive correlations between Italian secondary school learners' enjoyment and perceived competence pertaining to all four language skills, including L2 speakers' perceived competence in speaking. Furthermore, Dörnyei and Kormos (2000), Dörnyei (2002), and Kormos and Dörnyei (2004) report positive correlations between Hungarian school learners' task attitude (conceptualized as a combination of enjoyment and perceived usefulness of the task) and oral performance, in terms of number of words, turns, and arguments. In the area of reading comprehension, positive effects of enjoyment on achievement have been reported for university students (Dhanapala & Hirakawa, 2016) and for pre-vocational learners (De Milliano, 2013). Although research into the relationship between enjoyment and oral interaction is currently limited, these studies suggest that the existence of such a relationship is plausible.
So far, positive effects of WTC on the quality of interaction have not been reported in research. In a laboratory study, MacIntyre, Babin, and Clément (1999) found WTC to affect university students' decisions to engage in a difficult speech task, but not task achievement itself. Instead, task achievement was predicted by speakers' self-confidence. However, WTC does affect the frequency of language use in general (e.g., Hashimoto, 2002;MacIntyre & Charos, 1996;Yashima, Zenuk-Nishide, & Shimizu, 2004).  thus argue that higher levels of WTC increase opportunity for language practice and usage, which, in turn, is likely to facilitate the learning process. A similar point might be made for increasing levels of self-confidence and enjoyment. Positive emotions strengthen learners' awareness of language input, which allows learners to absorb the foreign language more effectively (MacIntyre & Gregersen, 2012), promote resilience in learners, and encourage learners to explore and play (Dewaele et al., 2018). In this light, this study explores whether development in learners' WTC, and self-confidence in and enjoyment of oral interaction positively affects their achievement in EFL oral interaction. MacIntyre et al. (1998) conceptualize WTC and self-confidence as distinctly situation-specific variables. This raises the question whether affect developed in one context of use supports learner achievement in another context of use. With regards to learnt skills, Lightbown (2008) posits that a sustainable transfer can only take place if the training context and the eventual interactional situation are closely matched. The same may be true for affective factors. For this reason, this study also explores whether affective factors explain achievement in EFL oral interaction both in trained and untrained contexts of use.

APPROACHES TO TEACHING ORAL INTERACTION
At present, the majority of commercially produced FL course books seem to adopt a largely controlled, form-focused approach to teaching oral interaction (e.g., Burns & Hill, 2013;Gómez-Rodríguez, 2010). In such an approach, both the focus of instruction and the focus of interaction activities lie on the development of language knowledge. Typically, explicit presentation of language forms is combined with activities such as dialogue repetition and filling in blanks. Ellis (2009) labels such activities as 'exercises.' Exercises engage learners in producing correct linguistic forms (i.e., those studied in class) but lack a clear communicative goal to be achieved. Successful performance is measured according to learners' grammatical correctness. Exercises are thus decidedly form-focused and allow for the internalization of linguistic forms. A much-used exercise for practicing oral interaction is the pre-scripted role play. It provides learners with pre-structured interactional situations in which speakers' roles are prescribed and known to both learners, and furthermore supply learners with language instructions (e.g., grammatical or lexical encoding, translation or responding to L1 content clues) which are prepared prior to interaction.
Ellis contrasts form-focused 'exercises' with 'tasks,' which are meaning-focused. These tasks engage learners in communicating content (meaning) and have a distinct communicative goal which the learners work toward. Task performance is successful when the communicative goal is achieved. Such tasks have the potential to evoke unpredictable interaction between speakers. An example of a meaning-focused task is the information gap task. In these tasks, each speaker holds part of the information (e.g., on separate role cards) necessary to complete a shared goal. Learners must thus interact with each other in order to complete the task, or solve a problem together. This generates unpredictable interaction.
Some course books supplement form-focused, pre-scripted exercises with information gap tasks. The instructional focus, meanwhile, remains predominantly on developing language knowledge. However, interactional ability not only hinges on language knowledge and speakers' ability to mobilize that knowledge in real time. Competent speakers also possess an array of interactional strategies that help them safeguard mutual understanding and address interactional problems when needed (e.g., Celce-Murcia, 2007;Dörnyei & Kormos, 1998). Very rarely do course books supplement form-focused instruction with instruction aimed at developing strategies that help solve interactional problems learners may come across when engaged in authentic interaction (Bueno-Alastuey & Luque Agulló, 2015; Faucette, 2001). These different instructional foci and task types will now be discussed, along with their potential advantages and disadvantages for fostering affect and achievement in EFL oral interaction.

EFFECTS OF TASK TYPE
Form-focused tasks have been known to play an important role in learners' automatization of language forms when moving from declarative to procedural knowledge through repeated practice (Anderson, 1982). Similarly, the negotiated interaction that results from information gap activities has generally been found to affect language acquisition positively (e.g., Doughty & Pica, 1986;Gass, Mackey, & Pica, 1998). Studies on the effect of task type on the development of FL oral interaction and on learners' willingness to communicate, self-confidence, and enjoyment of oral interaction are not available. However, there are some indications that favor the use of information gap tasks for fostering learner affect.
First, Dewaele, and MacIntyre (2014) and Dewaele et al. (2018) argue that challenging, unpredictable tasks that involve risk-taking and give learners a sense of autonomy can boost learners' levels of enjoyment. These demands seem better suited to information gap tasks than to formfocused tasks in which interaction is pre-scripted. The focus on goal achievement in information gap tasks is likely to challenge learners cognitively, and allows learners to take risks in trying out new language (Leaver & Kaplan, 2004), while the unpredictable course of interaction generated in such tasks provides learners with some autonomy over how to shape the interaction to achieve task goals.
Second, the pre-scripted nature of formfocused tasks allows learners to achieve the task goals (using a small set of language structures) with ease and in relative safety. Although this may support learners' WTC and self-confidence for using the FL within the classroom context, it also creates an illusion of mastery that underprepares learners for dealing with the unpredictability of real-world communication (Willis, 1996). This, in turn, may reduce learners' self-confidence and willingness to communicate in the FL in out-of-class situations. Information gap tasks, on the other hand, provide learners with the opportunity to practice solving interactional problems they come across during task performance (Pica, Kanagy, & Falodun, 2009), which may enhance learners' self-confidence and willingness to communicate in the FL. Foster (1998), however, warns that encountering too many interactional problems in information gap tasks can make learners feel unsuccessful and ineffective, which might have an adverse effect on learners' self-confidence and willingness to communicate using the FL.

EFFECTS OF INSTRUCTION
Competent speakers possess both language knowledge and a range of strategies that help address interactional problems they may come across during interaction. Interaction instruction can therefore either be form-focused, that is, aimed at developing the language forms necessary to fulfil specific language functions, or strategy-focused, that is, aimed at developing a set of self-supporting compensation and meaning negotiation strategies (e.g., Bygate, 1987;Dörnyei & Scott, 1995) and other supporting strategies such as attentive listening, responding to clarification requests, and erroneous interpretations of the message (e.g., Bygate, 1987).
The effects of form-focused instruction on EFL interactional ability are largely unknown (e.g., Norris & Ortega, 2000) and no studies are available that investigate how form-focused instruction affects learners' WTC, and self-confidence in and enjoyment of (E)FL oral interaction. There is some indication that strategy-focused instruction positively affects task effectiveness in EFL oral interaction (Lam, 2006). In this study, however, 'task effectiveness' was operationalized as assessors' ratings of general effectiveness and confidence in handling a task by a group of learners, thus causing overlap between ratings of ability and of affect. To our knowledge, the effects of strategyfocused instruction on WTC and enjoyment have not been investigated. Research into the effects of strategy instruction on self-confidence has produced mixed results. Lam (2006) reports that selfconfidence levels of low-proficiency learners of English increased after an eight-lesson intervention focused on explicit instruction in, amongst others, compensation, meaning negotiation, and time-gaining strategies. Cohen, Leaver, and Li (1996), however, report that 10-week strategiesbased intermediate programs for French and Norwegian led to an increase in self-confidence in only one of two speech tasks, and only for the students studying French. Here, communication strategies were taught as part of a larger set that also included (meta-) cognitive, social, affective, and performance strategies. In both studies, however, self-confidence was operationalized as confident (e.g., smooth, uninterrupted) task execution as observed by raters, an operationalization that seems more indicative of learners' speaking fluency than of self-confidence. In contrast, participants in Forbes & Fisher's (2018) study rated themselves on measures of confidence. After 6 weeks of explicit strategy instruction, they found the self-confidence of advanced learners of French to increase mainly on the basis of linguistic preparation strategies, and only partially on self-supporting strategies like selfcorrection, asking for clarification, and assistance. Overall, these studies report some positive effects of strategies-oriented instruction on learners' self-confidence.

THE PRESENT STUDY
This study focuses on pre-vocational learners in the Netherlands. Between 50-60% of adolescents between 12-16 in the Netherlands are enrolled in pre-vocational tracks (Centraal Bureau voor de Statistiek, 2016), which prepare learners for further vocational education and employment at the middle-management level (Liemberg & Van Kleunen, 1998), where they will use English for occupational purposes, that is, when interacting with non-Dutch third parties as part of their job. Developing strong EFL interactional skills is thus of pivotal importance for these pre-vocational learners. Despite the large number of learners enrolled in pre-vocational tracks, no large-scale classroombased research has been conducted that provides insight into these learners' development of oral skills in the EFL classroom. Nevertheless, learners' oral skills currently do not seem to meet expectations upon entry in further vocational programs, both in terms of ability, that is, accurate and fluent performance, and in terms of affect, that is, the confidence needed to interact in vocational situations (Jansma & Pennewaard, 2014). To address this issue, it is important to establish what type of instructional program would best foster pre-vocational learners' development of EFL oral interaction. In a related study, we evaluated the effects of 3 newly developed instructional programs on the development of pre-vocational learners' ability in EFL oral interaction. We found positive effects of these programs as compared to the effects of business-as-usual EFL instruction, with similar effects for each program. Considering the important role that affect plays in language learning, we now wish to find out which of these new programs best fosters learner affect (WTC, self-confidence, and enjoyment). Because affect also plays an important role in language achievement (Dewaele et al., 2018;MacIntyre & Gregersen, 2012), we additionally wish to explore to what extent development in WTC, selfconfidence, and enjoyment may explain learners' achievement in EFL oral interaction.
The three newly developed instructional EFL programs differed in instructional focus and type of task, that is, (a) a program that combined form-focused instruction and practice with pre-scripted interaction tasks (Form-Focused Interaction), (b) a program that replaced the prescripted interaction tasks with information gap tasks (Language-Directed Interaction), and (c) a program that combined these information gap tasks with interactional strategies instruction and practice (Strategies-Directed Interaction). These programs were all situated in a training context suited to the professional track that participants were enrolled in (Business & Administration studies).
Information gap tasks seem well-suited to the development of positive affect in FL oral interaction (e.g., Leaver & Kaplan, 2004;Willis, 1996), but effects have not been tested nor com-pared to effects of pre-scripted tasks. Furthermore, some positive effects of strategies-directed instruction on learners' self-confidence have been reported (Cohen et al., 1996;Forbes & Fisher, 2018;Lam, 2006), but these have not been compared to the effects of form-focused or languagedirected instruction. Moreover, it is not clear which of the three types of instructional approaches (form-focused, language-or strategiesdirected) has greater positive effects on WTC, selfconfidence, or enjoyment. We therefore posed the following research question: RQ1. Are pre-vocational learners' WTC, selfconfidence, and enjoyment of EFL oral interaction best fostered in a formfocused program, a language-directed interaction program, or a strategiesdirected interaction program? Prior research has shown that self-confidence and achievement in FL oral interaction are related (e.g., MacIntyre et al., 1997;Steinberg & Horwitz, 1986), but less is known about the relation between achievement in EFL oral interaction and learners' enjoyment or WTC.  argue that growth in WTC increases opportunity for language practice and usage, which is likely to facilitate the learning process. Growth in WTC might thus have a beneficial effect on achievement in EFL oral interaction. This has not been empirically tested in prior research, nor do we know whether growth in self-confidence and enjoyment is associated with higher achievement in EFL oral interaction. Furthermore, since affective factors are largely situation-specific (Mac-Intyre et al., 1998), it is important to establish whether such an association is dependent on learners' familiarity with the context of use. We therefore posed a second research question: RQ2. Does development in WTC, selfconfidence, and/or enjoyment explain achievement in EFL oral interaction, both in trained and untrained contexts of use?
To answer these questions, pre and post measures of learners' WTC, self-confidence, and enjoyment, and measures of achievement in EFL oral interaction were obtained. In our analyses, we controlled for individual differences in prior experience in EFL oral interaction, EFL vocabulary, and gender. Experience with oral interaction influences learners' oral performance (Trofimovich, Lightbown, & Halter, 2013) and helps control anxiety (Dewaele, Petrides, & Furnham, 2008). Vocabulary knowledge correlates both with oral performance (De Jong et al., 2012) and with WTC (MacIntyre & Legatto, 2010), and gender differences are related to oral performance (Krämer et al., 2014), the use of strategies (Oxford, Nyikos, & Ehrman, 1988), and levels of anxiety and enjoyment (Dewaele et al., 2016(Dewaele et al., , 2018.

Participants
Sixteen secondary schools in the Netherlands were invited to participate in this project. Eight schools expressed an interest in participation, three of which eventually accommodated the project, providing a total of 10 classes. One of the three participating schools was situated in a small town in the north of the country, and two were situated in a city near Amsterdam.
Participants (aged 14-15) were pre-vocational learners, in their third year of a 4-year prevocational Business & Administration program in the Netherlands (Grade 9). They were enrolled in the two lowest levels of pre-vocational education, roughly equivalent to level 2 in the International Standard Classification of Education (UNESCO, 2012). In accordance with local educational research guidelines and in close collaboration with the schools, all parents were informed about the study and the possibility of non-participation. One parent objected and their child was subsequently withdrawn from the study.
Data were collected from 156 participants. Nine participants did not complete the questionnaires and/or had attended less than 50% of the lessons and were subsequently deleted from the sample. We therefore report on a sample of 147 participants, 56.5% male. Of these participants, 60% were monolingual Dutch and 29.2% were multilingual speakers, 17.1% of whom reported English as one of the home languages. Of the sample, 10.8% had a non-Dutch language background.
Prior to this study, participants had received 2 years of compulsory EFL instruction at primary school and 2.5 years of compulsory instruction at secondary school. During these 2.5 years at secondary school, the participating schools timetabled an average of 120-135 minutes of English per week. All schools made use of course materials produced commercially in the Netherlands (New Interface and Stepping Stones), and the main language of instruction during lessons was Dutch. Teaching teams at all schools consisted of teachers trained to teach in the pre-vocational tracks.

Design
Between January and April 2016, a quasiexperimental design with pre-and posttest was implemented to assess the effects of three approaches to teaching EFL oral interaction on learner affect. Participants were randomly assigned within classes to one of three experimental conditions: Form-Focused (n = 48), Language-Directed Interaction (n = 46), and Strategies-Directed Interaction (n = 53). This resulted in dividing each class into three separate subgroups, each of which was taught as a separate group, in a separate classroom. Table 1 shows background information of the students in the three conditions. Conditions did not differ significantly with regards to students' home language, χ 2 (6) = 2.001, n = 147, p = .920, or gender, χ 2 (2) = .192, n = 147, p = .909. Additionally, analyses of variance (ANOVAs) indicated that learners between conditions did not differ significantly in terms of prior Experience in EFL interaction,

Interventions
To maximize chances that learners would experience the program as relevant to their own (future) lives, learners in all groups received instruction in a professional domain belonging to the Business & Administration program they were enrolled in, that is, in the role of hotel receptionist. In each of the three programs, identical sample dialogues were studied that modelled both the use of the targeted language structures and the use of target interaction strategies. Depending on the program's focus, these dialogues were accompanied by noticing and awareness activities aimed at either language structures or interaction strategies. Each program consisted of nine 40-45minute lessons that were taught within a 12-week time span. Learners were taught in groups of 5 to 8 learners. To control time on task, the programs contained similar numbers and types of activities, including 2 application tasks (on identical topics) per lesson, that is, a total of 18 application tasks over the course of the nine lessons. The programs differed in instructional focus and type of application task. Note. FFI = form-focused interaction; LDI = language-directed interaction; SDI = strategies-directed interaction advise a customer). These included modals of necessity, recommendation and obligation, connectives, adverbs of time, asking for preferences, comparatives, superlatives, and intensifiers. Activities included studying sample dialogues, noticing target forms in these dialogues, explicit rule presentation, and controlled practice activities, for example, matching clauses or conjugation activities. Application tasks were pre-scripted, form-focused tasks that guided participants through professional dialogues modelling the interactional encounter, and that provided language instructions, that is, grammatical encoding (see Appendix A for an example).
Language-Directed Interaction (LDI) Condition. Participants received exactly the same formfocused instruction and practice as the FF group, except that these were combined with information gap tasks (Appendices B, C, and D), designed to encounter a number of interactional problems in conversation, for example needing to explain a concept for which they lacked the vocabulary.

Strategies-Directed Interaction (SDI) Condition.
This condition explicitly taught interaction strategies considered helpful in addressing problems and maintaining mutual understanding in interactional encounters. These included compensation strategies (e.g., approximation, circumlocution, and exemplification), meaning negotiation strategies (e.g., indicating incomprehension and asking for elaboration, clarification, tempo adjustment, and repetition of the message), and audience awareness strategies (e.g., attentive listening, avoiding and addressing misunderstanding, and message alignment). Activities included studying sample dialogues, noticing target strategies in these dialogues, and explicit presentation and practice of said strategies. Application tasks were the same information gap tasks used in the LDI condition.

Procedure
Learners received 9 lessons. Groups were taught by 12 research assistants who had been recruited and trained specifically for this purpose. All assistants were university-educated, with a background in Education Studies, Pedagogy, or Psychology. Assistants were allocated to specific schools, where they each taught in one specific condition to prevent cross-conditional contamination. Lessons were taught in three separate classrooms, in parallel.
To maximize treatment fidelity, we trained the research assistants in their roles as teachers using teacher's guides that contained an explanation of the methodological approach, instructions for organizational and pedagogical conduct, and lesson plans containing lesson phasing, time limits, and protocols for instructions and explanations. Unannounced treatment fidelity checks carried out by the first author and reflection forms were filled out by the assistants after each lesson. No anomalies were detected. Reflection forms completed by the research assistants indicated that planned activities were delivered in all conditions, except that the second of two interaction tasks was not always implemented in lessons delivered in the LDI (89% and 86% implementation, respectively) and SDI (92% and 74%, respectively) conditions. This likely occurred because the content preparation required for information gap tasks is more effortful, and thus takes longer, than the linguistic preparation required for pre-scripted tasks.

Measures
Affect. To obtain measures of participants' affect in EFL oral interaction, identical questionnaires were administered before and after the intervention. Since affective factors are largely situated (MacIntyre et al., 1998, items in each scale contained a generally formulated item (e.g., 'I enjoy speaking English') and further distinguished between interacting in different contexts (school and leisure time) and with different audiences (peers and adults). To ascertain that items were understood as intended, the questionnaire was piloted with learners of a similar age, using Think Aloud Protocols. The questionnaire was in Dutch. Participants provided ratings on a Likert scale of 1-5 for the following components: 1. WTC: the extent to which learners are willing to engage in conversation using English, e.g., 'If my teacher asks me a question in English, I am happy to answer' (5 items, α = .76 and .77 for pre-and posttest). Items specific to measuring EFL oral interaction were selected and adapted from MacIntyre et al.

Achievement in EFL Oral
Interaction. Achievement in EFL oral interaction was measured with two interactional speech tasks after lesson 9. The first task was situated in the professional context in which learners had been trained. In this task, learners had to persuade a guest to buy a gift from the hotel gift shop. The second task was designed to check whether potential gains from instruction would transfer to an untrained, personal context. In this task, learners had to persuade a sibling's friend to buy their second-hand headphones. Both tasks were scripted speech tasks. This is an individual test in which one candidate's performance in EFL interaction is tested in interaction with an interlocutor, and in which the interlocutor's textual and interactional contributions are controlled through the use of scripts, thus standardizing both linguistic (complexity, register, style) and interactional (set points requiring the use of interactional strategies) challenges posed to candidates. See Van Batenburg et al. (2018) for a full discussion of task design, administration, and validation.
During the test session, participants carried out the tasks individually with a trained research assistant acting as interlocutor. Performances were video-taped and subsequently assessed by trained raters blind to condition on a Likert scale of 1-5 for the degree to which participants achieved the communicative goals set by the task. Each task was rated by a different rating team. Each team consisted of two raters. To establish interrater reliability, both these raters rated a set of 50 tasks randomly selected from the sample. They subsequently rated a set of about 60 tasks (50% of the remaining total) individually. Intra-class correlation coefficients (two-way random model, absolute agreement) were .81 for task 1 and .82 for task 2. Raters did not participate as teachers or interlocutors in this study.
Experience. In order to control for individual differences between learners that relate both to ability and affect in FL oral interaction, we obtained measures on learners' prior experience and vocabulary knowledge. On a Likert scale of 1-5, participants indicated how experienced they were in interacting in English both inside and outside of the classroom, both with adults and with peers, for instance, 'We often practice speaking in class'; 'In my free time, I often speak English with my friends. ' Dewaele and colleagues (2008) had found learners' anxiety levels to be related both to the size of learners' network of interlocutors and to the context of use (either in-class or out-of-class). In line with the aforementioned scales measuring affect, scale items for experience thus differentiated between audience and context. This newly developed scale consists of 6 items and proved reliable (α = .73). Deleting any of the items would not improve internal consistency. These data were obtained prior to the start of the intervention.
Vocabulary. Productive and receptive vocabulary are known to correlate highly. Meara & Fitzpatrick (2000), for instance, report a correlation of .841 between productive and receptive measures of vocabulary. We therefore chose to measure participants' vocabulary knowledge using the Peabody Picture Vocabulary Test (Dunn & Dunn, 2007), adapted for use in an EFL setting. Taking coverage in pre-vocational EFL course books as a selection criterion for determining item familiarity, 46 items were selected from sets 1 to 12 of the original test. The test covered different content areas (e.g., actions, sports, animals) and parts of speech (nouns, verbs, and adjectives). In a whole class setting, participants matched orally delivered vocabulary items with one of four pictures projected on a smart board by circling the correct number of the picture on their answer sheets (α = .85). The test was administered before the intervention commenced.

Analysis
The variables were examined for accuracy of data entry and distributions. No anomalies were detected. To establish if and how the type of EFL instruction participants received was related to the development of affect in EFL oral interaction, pre-and posttest mean scores for each of the three affective measures were calculated. Repeated measures analysis was subsequently conducted, analyzing how the development of the affective measures over time differed per condition. Because participants in this study were drawn from different classes, the data were structured hierarchically, meaning that independency between class and ratings on affective measures could not be assumed. Class was found to contribute significantly. For this reason, mixed model analysis was applied with Class added as random factor. Vocabulary, experience, and gender were covariates.
Hierarchical regression analysis was used to explore whether post-intervention levels of WTC, self-confidence, enjoyment, and instruction type contribute to EFL oral interaction, after controlling for the influence of pre-intervention scores, vocabulary, gender, and experience. Due to the hierarchical structure of the data, the contribution of Class to the regression model was investigated. This contribution was not significant. For this reason, unilevel analyses were conducted.
Preliminary analyses were conducted to ensure that the cases-to-predictor ratio was adequate and that there was no violation of the assumptions of normality, linearity, multicollinearity, and homoscedasticity. The sample size (N = 147) exceeds a minimum sample size of 90, which is sufficient according to the rule of thumb from Tabachnick and Fidell (2013) with regression models with 2 to 5 predictors. Visual inspection of normal P-P plots showed no violations of normality. A visual inspection of scatterplots showed no violations of linearity. In addition, linearity tests were significant with p values smaller than .001 for both outcome measures (Achievement in EFL oral interaction in Task 1 and Task 2). Bivariate correlations between predictor variables ranged from r = .35 to r = .80, indicating no strong multicollinearity or singularity (defined as r values > .9). In addition, collinearity tolerance values ranged from .33 to .80, i.e., not below the critical value of .1 (Tabachnick & Fidell, 2013). Finally, visual inspection of residuals with predicted measures plot showed no indication of heteroscedasticity.

Effect of the Intervention on Learner Affect in EFL Oral Interaction
Mixed model repeated measures analysis was used to establish whether pre-vocational learners' WTC, self-confidence, and enjoyment of EFL oral interaction are best fostered in a form-focused program, a language-directed interaction program, or a strategies-directed interaction program. To establish the effect of instruction type on development, interaction effects between instruction type and time were analyzed. Table 2 shows the means, standard errors, and effect sizes resulting from this analysis. Data were standardized, so that the parameter estimates of the independent variables could be interpreted as effect sizes (Cohen's d) while controlling for other parameter estimates in the analysis.
There were no significant main effects of condition on any of the three measures (Fs < 2.204,  With respect to Self-Confidence, both the main effect of time, F (1,144) = 6.910, p = .010 and the interaction effect of instruction type and time were significant, F (2,144) = 3.275, p = .041. For subsequent analysis of fixed effects, the alpha level was adjusted to α = .016 to correct for multiple tests. This analysis showed a significant increase from time 1 to time 2 for the SDI condition (b = .30, t = 3.52, p = .001), but not for the LDI condition (b = −.018, t = −.196, p = .845), nor for the FFI condition (b = .12, t = 1.36, p = .175). A small to medium effect (ES = 0.32) was found for the SDI condition compared to the LDI condition, b = .32, t (144) = 2.544, p = .012. Comparisons between the SDI condition and the FFI condition and between the LDI and FFI condition were not significant at α = .016. These results indicated that pre-vocational learners' WTC decreased from time 1 to time 2 regardless of instruction type, that their selfconfidence for EFL oral interaction developed significantly as a result of strategies-directed instruction, and that development in this condition differed significantly from the development in the LDI condition.

Development in Affect as a Predictor of Achievement in EFL Oral Interaction
To explore whether development in affect contributes to achievement in EFL oral interaction, regression analyses were conducted, both with a task set in the professional context in which learners had been taught (Task 1) and with a transfer task, set in a personal interactional context (Task 2). Since research so far had not revealed clearly defined dependencies between WTC, enjoyment, and self-confidence, each of these variables was investigated separately, resulting in six hierarchical multiple regression analyses. To take multiple testing into account, the alpha level was adjusted to .008.
To gauge the effect of development of WTC, enjoyment, and self-confidence, pretest scores of these measures were added as control variables in step 1 of each corresponding analysis. Adding the posttest scores as predictors therefore amounts to measuring to what extent the additional variance from time 1 to time 2 of that variable can explain the dependent variable. To control for individual differences likely to explain variation in EFL oral interaction, Vocabulary, Experience, and Gender were also entered at step 1 of each analysis. Experience did not contribute significantly to the analyses for Task 1, and was subsequently removed from the analyses for this task. Gender did not contribute significantly to Achievement in EFL oral interaction to the analyses in either task, and was therefore removed from all analyses. Entering Instruction Type in step 2 did not significantly improve any of the regression models ( R 2 's < .014, p's > .30), and was subsequently removed from all analyses. As explained before, posttest scores for each of the affective variables WTC, Self-Confidence, and Enjoyment were entered individually at step 3. This did not significantly improve the model for WTC in Task 1, R 2 = .000, p = .801 nor in Task 2, R 2 = .006, p = .284, nor for Enjoyment in Task 1, R 2 = .000, p = .992 or Task 2, R 2 = .000, p = .992. Entering posttest scores of Self-Confidence at step 3 did not significantly improve the model in task 2, R 2 = .269, p = .202, but it did significantly improve the model in task 1, R 2 = .042, p = .005. This regression model is displayed in Table 3. These analyses showed that, after controlling for the influence of pretest scores, vocabulary, and prior experience, development in self-confidence is related to achievement in EFL oral interaction in trained contexts, but not to achievement in new interactional contexts. Instruction type, WTC, and enjoyment did not predict achievement in EFL oral interaction.

DISCUSSION AND CONCLUSION
The first objective of this study was to gain insight into the effects of three different instructional programs on Dutch pre-vocational learners' affect in EFL oral interaction. We found that information gap tasks combined with strategies instruction (SDI) instigated a significant increase in learners' self-confidence. Form-focused instruction and practice, whether combined with prescripted tasks (FFI) or with information gap tasks (LDI) did not generate significant development of self-confidence. In all three programs, willingness to communicate decreased, while enjoyment of EFL oral interaction remained unchanged over time.
Both in the SDI group and the FFI group, self-confidence developed in a positive direction. In both of these groups, instructional focus was fully aligned with task type, that is, form-focused instruction and tasks in the FFI group and meaning-focused instruction and tasks in the SDI group. Such alignment seems to contribute positively to the development of self-confidence in oral interaction tasks. However, self-confidence did not develop in a positive direction in the LDI group. Both the LDI and SDI groups made use of information gap tasks to help learners gain experience in solving interactional problems during EFL interaction. As pointed out by Foster (1998), however, evoking problem-solving behavior through tasks is not without risk, because a substantial need for communicative repair may leave learners feeling incompetent. Indeed, the results of this study suggest that learners benefit most from information gap tasks if they have been introduced to useful interactional strategies during instruction and practice. Where such instructional support is missing, using information gap tasks does not increase learners' selfconfidence. These results are in line with the positive effects of strategy instruction reported by Cohen et al. (1996), Forbes & Fisher (2018), and Lam (2006).
Because little research is available on developing WTC through teaching, it is not immediately apparent how to explain that learners' willingness to communicate in English decreased over time. With the exception of MacIntyre et al. 's (2002) cross-sectional study on French immersion teaching, little is known about the timespan required for boosting levels of WTC, whether developing WTC is a linear process, whether it is conditional to the development of other affective factors, whether increasing WTC relates to individual factors such as age and proficiency level, and so on. MacIntyre et al. (2002) reported an increase in WTC over a period of 1 year, from Grade 7 to Grade 8, which might suggest that, unlike self-confidence, affecting change in WTC requires a lengthier intervention than the one conducted in this study. However, the authors also reported that the gains were not extended further in Grade 9, and suggest that to produce continuing gains in WTC, anxiety levels should be reduced further. If the development of one variable is indeed conditional to the development of another variable, it is conceivable that the gains of instruction on self-confidence in our study were not large enough yet to instigate an increase in WTC.
The programs did not significantly change learners' enjoyment of EFL oral interaction. MacIntyre et al. (2002) point out that motivation for school work generally tends to decrease during adolescence (cf. Sigelman, 1999). This corresponds with Dewaele et al.'s (2018) findings that foreign language enjoyment significantly drops around the age of 14-15. These authors furthermore report that low-intermediate learners' enjoyment levels are significantly lower than those of high-intermediate or advanced learners. Thus, not only age but learner level seems to be an important factor that correlates with enjoyment. In this light, we may need to consider the possibility that the enjoyment of FL oral interaction is a relatively stable feature for low-proficiency adolescent learners, such as those who partook in this study, and is thus less malleable through instruction (cf. Gardner & Tremblay, 1994).
Finally, these findings might be explained as a result of conducting classroom-based research, in which both educational goals and research aims must be balanced. Aiming to control as many variables as possible in the study resulted in adopting an identical lesson structure for each of the 9 lessons in each of the three experimental programs. At task level, the design recognized the importance of unpredictability, autonomy, challenge, and risk-taking to boost levels of enjoyment (Dewaele & MacIntyre, 2014, Dewaele et al., 2018 by juxtaposing pre-scripted interaction tasks with information gap tasks. The absence of variation in the overall lesson structure, however, may have rendered the lessons largely predictable. This may have led to boredom on the part of the learners, which might have hampered their development of enjoyment and, in turn, WTC. Because affect also plays an important role in language achievement (Dewaele et al., 2018;MacIntyre & Gregersen, 2012), we additionally explored whether development in WTC, selfconfidence, and enjoyment explains learners' achievement in EFL oral interaction and, if so, whether it does so both in trained and untrained contexts. Our analyses showed that instruction type and development in enjoyment and WTC do not significantly predict achievement in EFL oral interaction, but that development in selfconfidence does explain achievement to some extent. This is in line with MacIntyre et al.'s (1999) study in which WTC was found to affect students' decisions to engage in a difficult speech task, but only self-confidence was found to predict task achievement.
Self-confidence only explains achievement in the interaction task that matches the professional context in which learners were trained. In other words, the self-confidence gained in lessons situ-ated in a hotel context helped learners perform better in the hotel task, but did not help them perform better in the task that was situated in a different, personal context. These results substantiate the notion that self-confidence is situationspecific (MacIntyre et al., 1998). They furthermore suggest that gains of instruction do not automatically transfer from one context to another. In a related study that evaluated the effects of form-focused, language-directed, or strategies-directed interaction instruction on the development of learners' ability in EFL oral interaction, we found similar results (Van Batenberg et al., 2019). This led us to conclude that a sustainable transfer of learnt skills can only take place if the training context and the eventual language use situation are closely matched (cf. Lightbown, 2008). The results of the current study seem to suggest that this may not only be true for the development of ability, but also for the development of learner affect.
The current study shows that instruction type does not contribute significantly to achievement in EFL oral interaction. This confirms the results from the aforementioned study into developing learner ability in oral interaction, where we found positive effects of the three programs as compared to the effects of business-as-usual EFL instruction, but not differential effects between these three types of instruction (Van Batenburg et al., 2019). Thus, although instruction type does not directly influence learner achievement, it does influence the development of learner affect in EFL oral interaction: Self-confidence increases most as a result of strategies-directed instruction. Furthermore, development in self-confidence predicts learners' achievement in EFL oral interaction in trained contexts.

Limitations and Suggestions for Future Research
This study has shown that strategies-directed interaction instruction positively affects learners' self-confidence. Differential effects between conditions on other measures of affect were not found. In all three programs, learners' enjoyment of oral interaction remained stable, while learners' WTC decreased somewhat. As mentioned previously, more insight into the development of WTC in classroom-based learning is needed to fully explain these findings. Possibly, a lengthier intervention or more varied lesson design may be needed to incite more substantial, positive change in learner affect in EFL oral interaction. Another limitation is that in the lessons that incorporated information gap tasks (Language-Directed and Strategies-Directed Interaction), the second of two information gap tasks was not always implemented. It is possible that the content preparation required for information gap tasks is more effortful than the linguistic preparation required for pre-scripted tasks, and more time is needed to accomplish this. To gain a more in-depth understanding of the potential merits of instructional focus and task type, future studies should compare fully implemented programs. Despite these limitations, however, this study has shown that pre-vocational learners' self-confidence can be affected through teaching.

Implications for Practice
For many EFL learners worldwide, English is a compulsory school subject that is commonly taught with the use of EFL course books. Both within and outside the Dutch context in which our study was situated, the interaction practice offered in these course books is largely form-focused, combining form-focused instruction and practice with form-focused application tasks or, occasionally, with an information gap task. Strategy instruction, on the other hand, is largely absent from commonly used course books (e.g., Bueno-Alastuey & Luque Agulló, 2015). This study has shown that a form-focused approach, whether combined with pre-scripted form-focused tasks or with information gap tasks does not significantly increase self-confidence in EFL oral interaction, but that the use of information-gap tasks combined with strategies instruction does. On this basis, practitioners who aim to enhance their learners' self-confidence in oral interaction could increase the use of information gap tasks, while supporting task performance with interactional strategy instruction. The study furthermore provides a first indication that achievement in EFL oral interaction is to some extent predicted by growth in self-confidence. This suggests that, in aiming to improve learners' EFL oral interaction, it is worthwhile for practitioners to address self-confidence in their oral interaction lessons.