While general vocabulary acquisition is crucial in second language learning and EFL proficiency, there has been a growing focus on acquiring multiword vocabulary units alongside single words in recent years (e.g., Boers, 2021; Crowley et al., 2023; Wolter, 2020). Learning vocabulary in formulaic chunks—sequences of commonly used words, which may or may not carry idiomatic meanings when combined—has been proposed to improve contextual vocabulary acquisition, leading to enhanced fluency and more precise language production (e.g., Boers, 2021; Haugh & Takeuchi, 2023; Koh, 2020; Simpson-Vlach & Ellis, 2010). However, distinctions have been made between general collocations of words that co-occur without creating new and unexpected meanings from the individual parts (e.g., “relate to,” “show a trend”), and units of individual words that combine to produce meanings beyond their literal interpretations, such as phrasal verbs (e.g., “look around,” “throw away”) and idiomatic expressions (e.g., “kick the bucket,” “be in someone’s shoes”; Crowley et al., 2023; Sag et al., 2002).

Phrasal verbs are typically defined as combinations of a verb and a particle, which can either be adverbial or prepositional, that come together to form a new meaning, often related to motion, change, aspect, or idiomatic expression (Gardner & Davies, 2018; Haugh & Takeuchi, 2023; Spring, 2018, 2019). A number of studies indicate that EFL learners with verb-framed L1s, such as Japanese and Korean, often struggle learning these units due to a lack of awareness in how the verb and particle combine to create new meanings (Haugh & Takeuchi, 2023; Spring, 2018, 2019; Yasuda, 2010). Phrasal verbs have also been gaining increased attention in EFL teaching and learning studies because of their prevalence in English and polysemy, making them essential, but very difficult to acquire for many EFL learners (e.g., Al-Otaibi, 2019; Birdsell & Kavanagh, 2023; Rudzka & Ostyn, 2003; Spring, 2018, 2019; White, 2012; Yasuda, 2010).

Idiomatic expressions are not clearly defined, but share some similarities with phrasal verbs, especially in their figurative interpretations. While the term “idiomatic” encompasses various linguistic phenomena like metaphors and similes, this paper specifically focuses on non-compositional expressions, which had distinct literal and figurative meanings (Titone & Connine, 1999). For the sake of clarity, we exclude phrasal verbs, which consist of a single verb and a single particle (Gardner & Davies, 2018), from the category of idiomatic expressions to create completely separate categories. Although there is less research on the impact of L2 learning on idiomatic expressions in general, Cooper (1999) highlights the challenges that learners face with them in EFL. Additionally, scholars such as Haugh and Takeuchi (2023) and Yasuda (2010) have observed that the idiomatic meanings of phrasal verbs pose particular difficulties for EFL learners. However, idiomatic expressions themselves have not received nearly as much attention in EFL teaching and learning studies as phrasal verbs, given the large number of papers on learning the latter, specifically, and the rather sparse number dedicated to learning the former.

Finally, it is important to note that possessing a deeper understanding of multiword vocabulary, which encompasses phrasal verbs and idiomatic expressions, correlates with higher overall EFL proficiency (e.g., Boers, 2021; Kim, 2019), as well as higher scores on standardized tests (Crowley et al., 2023). Therefore, while multiword vocabulary may not receive as much emphasis as single-word vocabulary, it remains a crucial aspect of language learning that warrants investigation.

Traditional, deliberate acquisition of vocabulary, which may come in the form of word lists in textbooks or stacks of paper flashcards, have received criticism because they are monomodal and often decontextualized (e.g., Barcroft, 2020). However, a number of studies have suggested that online vocabulary learning tools can be used to acquire L2 vocabulary, specifically in the L2 context. In particular, technological advances in online flashcards have allowed for various practice modes that provide can provide contextualized multiple-choice questions (e.g., Nakata, 2020; van den Broek et al., 2023), gap-fill activities (e.g., Strong & Boers, 2019; Strong & Leeming, 2024), and technologyinfused speaking and listening practice (e.g., Spring & Takeda, 2023). One particular advantage suggested to using digital flashcards over more traditional paper-based study include increased motivation. For example, Ying et al. (2021) found that young learners showed more interest in studying after implementing online flashcards as a part of their foreign language class. Similarly, Spring and Takeda (2023) found that university EFL learners found flashcards motivating but noted that adding point systems to the flashcards improved their motivational impact. Another advantage to using digital flashcards is that they can encourage learners’ item recall when answers are shown on a delay (e.g., van den Broek et al., 2023). Specifically, van den Broek et al. (2023) have shown that when learning a foreign language, providing a cue followed by correct answers or choices after a short pause increases the odds of students remembering the words in the future. While paper-based flashcards or covering answers in a book could also provide this sort of practice, teachers do not have control of student behavior, so cannot ensure that students are practicing as efficiently as possible. Conversely, digital flashcards allow for delayed actions, forcing students to undergo recall before answers are shown (e.g., Strong & Leeming, 2023). Finally, some studies have suggested that the interactive nature of digital flashcards can make them more engaging. With paper-based practice, students generally only have to turn a page or shuffle a paper flashcard. However, digital flashcards can be modified to force students to click on correct answers (e.g., Nakata, 2020; van den Broek et al., 2023), write answers (e.g., Spring & Takeda, 2023; Strong & Leeming, 2024), or even watch and interact with videos (Hamagami et al., 2024). This is important because it is also widely thought that increased student engagement leads to higher motivation to learn, especially in online learning environments (e.g., Martin & Bolliger, 2018).

However, it should be mentioned that the effectiveness of these online tools seems to depend on how much they activate memory recall (van den Broek et al., 2023), a learner’s initial and corrected guesses (Strong & Boers, 2019), and the practice mode used (Spring & Takeda, 2023). Furthermore, though Strong and Boers (2019) and Strong and Leeming (2024) show the effects of such online tools on the learning of phrasal verbs, specifically, most studies to date have focused on the using of these tools for learning single word vocabulary items (Nakata, 2020; Spring & Takeda, 2023; van den Broek et al., 2023) and there are no studies that we are aware of that explore the use of such tools to learn idiomatic expressions that are not phrasal verbs. This is important because idiomatic expressions are often more likely to be used in spoken contexts (Zareva, 2016), suggesting that text-based practice methods such as gap-fill activities and multiple-choice flashcards might be less effective for acquiring these items for use in oral contexts, where they are particularly relevant.

In order for vocabulary to be truly acquired, learners should be able to process the word in reading, writing, listening, and speaking modes. For this reason, Nation (2020) recommends vocabulary learning take place across these multiple skills. However, in the context of learning a foreign language, learners frequently encounter vocabulary in reading and writing situations, primarily because these skills are often taught using textbooks, and in the Asian EFL context, many learners often attempt to memorize vocabulary through paper-based word-lists such as vocabulary books (e.g., Yoshitomi et al., 2006). To give students more access to auditory and oral vocabulary practice, there has been a push from EFL practitioners to incorporate videos, audio, and other multimedia resources into learning curriculums and teaching practices (e.g., Kaku, 2018; Kim, 2019; Koh, 2020; Spring, 2019, etc.). Theoretically, this is very important to compliment the more traditional reading and writing focus often found in Asian EFL contexts. However, despite these efforts, it has been difficult to point to the exact benefit of learning vocabulary through multimedia in a controlled learning context.

One reason it has been difficult to see the specific benefits of specific speaking activities on specific multiword vocabulary acquisition is also an issue with many flashcards and online tools; they tend to offer mostly reading or writing skills. For example, Quizlet flashcards often simply allow learners to pair L1 meanings to L2 words, but are generally based on visual input, making them reading focused (e.g., Kaku, 2018). However, studies such as Strong and Leeming (2024) and Nakata (2020) have also suggested that similar online tools can be created in which students write their answers. Furthermore, Takeda (2023) also suggested flashcards that include speaking and listening questions (based on text-to-speech listening and automatic speech recognition technology) may promote language acquisition. Spring and Takeda (2023) conducted an initial study that compared how multimodal flashcards that encouraged reading, writing, speaking, and listening could be utilized in the EFL classroom and verified that such multimodal usage helped students to obtain higher scores on subsequent quizzes. Furthermore, they discovered that using a variety of modes, and writing in particular, when studying was more effective than multiple-choice reading-based practice alone. However, their study focused on derivational single-word vocabulary and did not include a delayed test in the target context, so it is still unclear if the same trend will be seen for multiword vocabulary, which are more common in oral registers. Furthermore, it is also not yet known how multimodal practice will affect students’ performance on a delayed test that requires contextual listening and understanding of the multiword vocabulary units.

Based on the previous studies introduced above, there is reason to believe that learning multiword vocabulary units could have a discernible impact on learners’ general EFL proficiency and that learning them through online tools could facilitate this learning. However, it is still not yet clear what the impact of learning through multiple modes is on the ability to (1) remember the units initially, or (2) listen to and comprehend these multiword vocabulary units in conversations later. In order to investigate this in detail and uncover the role of various modes in EFL multiword vocabulary acquisition, we pose the following research questions:

1) Which modes of practice (reading, writing, speaking, and listening) had the greatest impact on initial retention of multiword vocabulary unit memorization?

2) Which modes of practice (reading, writing, speaking, and listening) had the greatest impact on delayed ability to listen to and comprehend multiword vocabulary units in conversations?

A total of 229 L1 Japanese EFL learners who were first-year university students agreed to participate in this study. They were provided with informed consent and allowed to withdraw participation in the study at any time in accordance with the ethical review board at the authors’ institution. The participants had taken the TOEFL ITP® test of general academic English proficiency one month prior to the study and exhibited a wide range of scores with most students being around a high CEFR A2 level to a low CEFR B1 level; 367-653 (M = 500.3, SD = 39.53).

The students participated in the activities described in this study as part of their university designated general English education classes. Students took one class focused on increasing listening and speaking skills taught by one of the two authors, and one class focused on reading and writing skills taught by another teacher. Students in the authors’ classes had to learn lists of phrasal verbs and idiomatic expressions as designated by their curriculum and course textbook, Pathways to Academic English, 4^th edition (Spring & Scura, 2023). The expressions for the textbook were selected from Martinez and Schmitt (2012) and Simpson-Vlach and Ellis (2010) based on co-occurrence in both lists and suitability to the definitions of phrasal verbs and idiomatic expressions provided in Section 2.1. Learning these multiword vocabulary units was included in the speaking and listening centric class due to the fact that the selected expressions are argued to be common in spoken academic contexts.

Classroom time for both authors consisted of using worksheets condoning pair and group speaking exercises, listening to audio files or watching videos meant to contextualize the target vocabulary, and later assigning weekly homework assignments meant to practice the target vocabulary and skills through reading, writing, listening, and speaking.

In order to help students learn and memorize the lists of phrasal verbs and idiomatic expressions, the authors created a multimodal flashcard system in which students could learn the multiword vocabulary units in L1-L2 matching, reading, writing, listening, and speaking modes. The model and technology mirrors that of Spring and Takeda (2023) and an example is shown in Figure 1. The L1-L2 matching flashcards presented learners with the definition of the target multiword vocabulary unit in their L1 and four randomized multiple-choice options, one of which was the target unit. The reading flashcards were nearly identical, but learners were prompted with a sentence containing a blank in which one of the multiple-choice options would be acceptable. The writing flashcards also contained a sentence with a blank, but then asked students to write the missing piece of the phrase or expression, following Strong and Boers (2019). The speaking and listening flashcards present the users with a short phrase or sentence containing the target multiword vocabulary unit in context as well as “hear” and “speak” buttons. Upon clicking the “hear” button, students heard the phrase or sentence pronounced by in-device text-to-speech technology. Upon clicking the “speak” button, the user’s in-device automatic speech recognition application programming interface (API) is utilized to record the user’s voice as they attempt to pronounce the phrase or sentence. The flashcard then analyzes the text output from the API and checks for the multiword vocabulary unit, considering inclusion of the unit as a “correct” pronunciation.

Students were provided with login credentials so that their data could be tracked and managed by their teachers. Each multiple-choice click, writing attempt, listening click, and speaking click was recorded and the number of attempts and correct usages were available for students to look at and for teachers to check. For this study, the authors asked students to use the flashcards and do a minimum of 30 total attempts before weekly quizzes on sets of the multiword vocabulary units but did not dictate the mode that students were required to use. The authors downloaded their students’ data just before the quizzes and once again before their final tests at the end of the semester.

Students took two quizzes of phrasal verbs, each containing 16 phrasal verbs, and four quizzes of idiomatic expressions, each containing between 11 and 20 expressions. The quizzes were administered weekly in the classroom through Google Forms, for a total of six weeks. The quizzes were created multimodally to promote the acquisition of the units in a variety of contexts (Spring & Takeda, 2023; Uchihara, 2022). Specifically, each quiz contained four question modes containing five questions each:

1) Listening questions in which students had to listen to a short phrase or sentence and choose the multiword vocabulary unit that they heard from four options (multiple-choice)

2) Meaning matching questions in which students had to choose the correct meaning of the target multiword vocabulary unit (multiple-choice)

3) Context matching questions in which students had to read a sentence containing a blank and choose the multiword vocabulary unit that best completed the sentence (multiple-choice)

4) Writing questions in which students had to read a sentence containing part of the multiword vocabulary expression and then fill in the missing piece of the expression (written, free answer)

Later, during the final class of a 15-week semester, the authors gave a final test to their students that contained 12 short conversation questions and 4 long conversation questions. All of these items were designed to test students’ delayed ability to listen to the multiword vocabulary units studied throughout the course and comprehend their meaning in the given context, with half of the questions were designed to test the comprehension of phrasal verbs in context and the other half were designed to test idiomatic expressions. The test also included other questions to test different skills not relevant to this study, but we isolated the data for these questions to obtain a measure of how well students had acquired the ability to listen to phrasal verbs and idiomatic expressions and understand them in context.

For this study, the total number of multimodal flashcard clicks were used to gauge the amount of practice undertaken in each mode. Additionally, we assessed students’ average quiz scores to measure their initial memorization of the multiword vocabulary items. As mentioned above, we considered students’ final test scores to evaluate their overall delayed ability to listen to and comprehend the target items in context. Lastly, we used students’ TOEFL ITP® test scores taken before the learning had begun to represent students’ initial general English proficiency.

To access the impact of different practice modes via flashcards on learning, we employed Mizumoto’s (2023) method of multiple regression with random forests the determine significance, along with dominance analysis with relative weights for relative importance. This statistical method was used due to random forests being more accurate than p values in multiple regression models for indicating predictive significance, and because dominance analysis helps counteract co-variance when uncovering the relative importance of each variable in the model. The first set of regression models used average quiz scores as the predictor variable and the second set used final test scores. Both sets of models included the four practice modes and TOEFL ITP® scores as the dependent variables. The TOEFL ITP® tests scores were included to account for initial English proficiency. We expected TOEFL ITP® scores to be the largest predictors of quiz and test scores as the learning phrase took place over a matter of just seven weeks. We then considered any other variables confirmed as significant by the random forests to have had a significant impact on learning. Furthermore, we used the relative importance metrics as a way to check which practice modes had the most impact on overall learning. We then make generalized comments based both on these analyses, the descriptive data, and contextualize the results based on other similar studies.