Comparison of Learning Content Representations to Improve L2 Vocabulary Acquisition Using m-learning

Previous works reveal that there is potential in the use of mobile devices as a useful tool to help learn new vocabulary and it allows the learning materials can be displayed with different Learning Content Representation (LCR) types. Nevertheless, there are no conclusive results about which LCR type is better to improve L2 vocabulary acquisition using m-learning. The purpose of this study was to explore the effects of improving the students’ academic performance using two LCR types that promote different learning strategies in a vocabulary learning task. A mobile application with two LCR types was implemented and tested: the SRL type based on a self-regulated learning strategy, and the NSRL type based on a non-self-regulated learning strategy. Quantitative analyses indicated that there is a relationship between the LCR types and the word recall on vocabulary learning. The SRL type seems to be effective in improving the students’ learning abilities and students in this experimental group exhibited significantly better academic performance. The results are meant to draw that LCR-SRL types must be used in m-learning to L2 vocabulary acquisition. Future studies should focus on how to integrate these strategies in m-learning and understand their effects on improving students’ vocabulary acquisition. Plain Language Summary Exploring mobile learning approaches for better second language vocabulary learning Mobile devices have emerged as promising tools for facilitating vocabulary learning, offering the flexibility to present learning materials in various formats known as Learning Content Representation (LCR) types. However, the question of which LCR type is most effective in enhancing L2 vocabulary acquisition in mobile learning remains unanswered. This study aimed to investigate the influence of two distinct LCR types, each promoting different learning strategies, on students’ academic performance in vocabulary learning tasks. To accomplish this, a mobile application featuring two LCR types was developed and tested. The first LCR type, Self-Regulated Learning (SRL), encouraged learners to employ self-regulation strategies, while the second LCR type, Non-Self-Regulated Learning (NSRL), guided learners through a non-self-regulated approach. Quantitative analyses of the data revealed a significant relationship between LCR types and word recall in vocabulary learning. Notably, the SRL-based LCR type demonstrated effectiveness in improving students’ word recall abilities, resulting in significantly better academic performance for the experimental group. These findings emphasize the importance of incorporating LCR-SRL types in mobile learning environments to enhance L2 vocabulary acquisition. Future research should delve deeper into the integration of these strategies within mobile learning platforms and explore their broader impact on students’ vocabulary acquisition. This study underscores the potential for optimizing mobile learning tools to foster effective language acquisition strategies in educational contexts.


Introduction
The Use of Mobile Learning to Learn New Vocabulary Today the use of technology in education is common, for example, virtual laboratories, computer simulations, the use of mobile devices, online teaching, interactive whiteboards, among others.Specifically, the development of the Internet and wireless networks has led to the use of mobile devices as educational technology tools.M-learning is an approach that can be complementary of both traditional learning as well as e-learning (Kumar Basak et al, 2018).M-learning allows learners to interact with their learning resources from an ubiquitous perspective (J.D. Clark, 2007;Levı´-Orta et al., 2020;Sevillano & Va´zquez-Cano, 2015).Kothamasu (2010) established five basic characteristics for an adequate development of m learning: (1) portable, (2) social interaction, (3) sensitive to the context, (4) connectivity, and (5) customized.Furthermore, Mohanna (2015) stated that m-learning can be enriched with different formats such as games, short messages, quizzes, and multimedia contents.In this sense, Pollara (2011, p. 67) stated that m-learning ''provides the ability to gather data unique to the current location, environment, and time real and situated.'' Research has been done on the advantages of the use of m-learning as an educational tool.He et al. (2012) demonstrate that the use of m-learning in combination with videos helps to improve the students' performance in a chemistry course.Taleb et al. (2015) present a study to analyze the effect of m-learning on teaching mathematics.Their results demonstrate that there is a significant and positive relationship between the use of mlearning and a student's participation in class.In addition, teachers believe that m-learning has a positive effect on motivating the students toward mathematics.In general, these previous studies claim that m-learning can be used as an effective tool for instruction instead of traditional methods.In this sense, the meta-analysis of Yu and Trainin (2022) indicated that L2 vocabulary learning assisted by technology across various conditions was more effective than instruction without technology.Although the effectiveness of technology in learning vocabulary could be affected by different variables, among which it can be highlighted: type of instruction, type of assessment adopted in the study, participants' grade level and their native language, and type of technology.
Some authors have studied the impact of the use of m-learning in comparison with traditional learning methods to learn a foreign language.For example, C. M. Chen et al. (2006) propose an intelligent m-learning system for supporting English learning.Their system recommends English news articles to students based on students' reading abilities.The new words are automatically discovered and retrieved from different articles according to the students' comprehension and their ability for enhancing vocabulary learning.Evans (2008) describes a study of the effectiveness of m-learning for teaching undergraduate Business and Management students in an Information and Communications Technology course.His results demonstrate that students believe that the use of m-learning is more effective as a teaching tool in comparison with the use of traditional methods such as reading a textbook.The results also suggest that students find the use of mobile devices to be effective, efficient, and engaging for learning.W. H. Wu et al. (2012) present a meta-analysis approach to systematically review the literature regarding mobile learning.Their results indicate that 86% of previous studies reported favorable advantages as a result of using mlearning in class.Klimova (2019) demonstrates that students who used m-learning significantly improve their performance for learning new English words in comparison with students who did not use the technology.Moreover, Hao et al. (2019) evaluate the effectiveness of an m-learning application and their results show that the use of the technology enhances the learning of a foreign language.
In this way, m-learning is becoming a useful tool for learning, including foreign language learning (Gangaiamaran & Pasupathi, 2017;Heil et al., 2016;Kacetl & Klı´mova´, 2019).Kukulska-Hulme (2016) mentions that the use of m-learning to learn a second language contributes to the development of the four language skills (listening, reading, writing, and speaking).
In the last 10 years, different quantitative studies have been published to investigate the influence of technology on the acquisition and improvement of the use of vocabulary through mobile and ubiquitous learning, microblogging and gamification (Dodigovic, 2013).Several studies have demonstrated that the use of technology improves L2 vocabulary acquisition.For example, Zhao (2013) with a sample size of 419 and found a large effect size, Cohen's d = 1.12.Taj et al. (2016) meta-analysis is one of the latest in the field, which included 13 studies, n = 813 and a large effect size of d = 0.80.And recently, Yu and Trainin (2022) with 2,511 participants and an overall effect size (d = 0.64) for using technology to learn L2 vocabulary.
For new learners, one of the main activities to improve their communication skills is the learning of new words.Q. Wu (2015) developed an Android smartphone app and investigated its effectiveness as an educational tool in helping English with college students to learn new vocabulary.He tested the app's effectiveness with a test group (using the app) and a control group (using a traditional method).His results showed that students from the test group significantly outperformed those using the traditional method; the students in the test group recognized 88 words more than the students who did not use the app.Rezaei et al. (2014) reported the use and effectiveness of mobile applications in English vocabulary learning.They analyzed the academic performance of intermediate-level English students before and after using m-learning, and their results claim that the use of m-learning helps increase the learning of vocabulary and students' confidence.
In general, previous works reveal that there is potential in the use of m-learning to help learn new vocabulary in comparison with traditional methods and learning materials had been supported using different learning contents, mainly visual and verbal.In this sense we share with Yu and Trainin (2022) that the challenge is to find the criteria to select appropriate vocabulary learning apps, turn them into effective tasks for L2 learners, satisfy L2 learners' different needs, and develop selfregulated strategies.
Learning Content Representation Types Weinstein and Mayer (1986) define learning strategies as specific behaviors and thought processes used by students to facilitate the acquisition, storage, or retrieval of information.Students use verbal and visual systems to memorize new vocabulary.The verbal system deals with writing words, while the visual system deals with mental images to relate it with the word (C.J. Chen & Liu, 2011).According to (J.M. Clark & Paivio, 1991), the verbal route encodes linguistic information in all its forms, whereas the visual route encodes images.When the inputs of the two routes overlap, encoding and retrieval improve.For this purpose, students need to reinforce the referential connections between the two codes in order to allow operations such as imagining words to reinforce input and accurate retrieval of information.
In general, one of the main problems to learn vocabulary is that students have problems in lacking learning strategies for memorizing new words.When a subject wants to learn a new language, there are some skills that must be trained with the purpose of having successful communication.According to Uso-Juan and Martinez-Flor (2006), foreign language learning is composed of four main skills (reading, listening, speaking, and writing), and two sub-skills (vocabulary and grammar).Nevertheless, for the purpose of supporting the main skills, the sub-skills must be trained, and they should be offered in context.Consequently, the learning strategies must ensure that the learners train each sub-skill to learn new vocabulary and/or understanding the grammar structures.
Specifically for vocabulary learning, different studies have been published that relate gamified apps with greater autonomy and vocabulary learning.Mainly by linking a series of fundamental words for the development of communicative competence and whose repetition and implementation cannot be guaranteed in faceto-face classes.The scientific literature has identified that students must learn between 2,000 and 3,000 basic words to participate in conversations and be able to read texts in English with a minimum of knowledge.In addition, the monitoring of the students' assessment of these practices can be favored by the learning analytics that allow the integration of these apps.In this sense, the metaanalysis by Zou et al. (2021, p. 22-23) concluded: ''(1) digital games promote effective vocabulary learning; (2) interactions in game environments are conducive to vocabulary learning; (3) game-embedded multimedia facilitates vocabulary learning; and (4) over-specified vocabulary information is better than isolated or minimally specified information.''Previous studies (Rezaei et al., 2013(Rezaei et al., , 2014;;Q. Wu, 2015) have demonstrated the advantages of the use of mlearning to improve the students' skills to memorize vocabulary in comparison with traditional teaching methods.The main advantage is that m-learning does not only improve academic performance, but it also promotes students' engagement (Klimova, 2019) and promotes an active learning scenario through a multimodal interaction (known as multimodality) between students and mobile devices (Seta et al., 2008).
M-learning provides user-friendly support with different modalities to access and interact with the learning content, students can see, hear and/or touch to memorize new words using different learning strategies.Thus, learning materials can be displayed with different Learning Content Representation (LCR) types (C.J. Chen, 2014;Sadler-Smith, 2011).Using different LCR types, learners can see images in the context of the meaning of a word, listen to the pronunciation of the word, and/or touch the screen of the device to interact with the mobile application.
N. S. Chen et al. (2008) mentioned that one disadvantage of m-learning is the screen size of mobile devices.This characteristic should be considered for the selection of learning information that will be displayed by the LCR type.Al-Samarraie et al. (2016) explored the effect of different LCR types on learners' performance.Their results demonstrated that using an LCR type with a simplified and easy-to-read display will help increase students' motivation for continued use of the learning application.Their results suggested that designers of learning systems should take into consideration the basic requirements of LCR types in terms of clarity and simplicity.
In this sense, personalized learning and the process of self-regulation are related.Ingkavara et al. (2022) recent work on personalized learning emphasizes the importance of tailoring teaching to the needs and preferences of individual students.By providing different representations of content, students can choose the mode that best suits their cognitive style, thereby supporting selfregulation process.Learning Content Representation (LCR) empowers students' SRL by adapting their learning strategies to monitor and enhance their learning (C.-M.Chen, 2009;Zheng et al., 2019).Likewise, recent studies have shown a relationship between the improvement of SRL processes when a technology-enhanced learning system provides different representations that allow the continuous practice of a content or competency (Berglas-Shapiro et al., 2017;Chang et al., 2022).
Another key element of self-regulated learning is metacognition and choice of content representation.When the meta-cognitive processes are fed back with different resources, especially those that allow the contents to be represented in different ways, there is an increase in adaptive learning (Cortese, 2022).Metacognition involves being aware of one's cognitive processes.By presenting content in a variety of formats, educators can encourage students to engage in metacognitive processes, such as reflecting on their learning preferences and strategically choosing the most effective ways to learn (Fan, 2003).In this sense, Barcroft (2009) showed that there is a positive correlation between the number of strategies used and the learning of foreign language vocabulary.In his study he observed that students obtain better grades when they use mnemonic techniques and different models of representation in picture or figure format; a much greater effect than strategies based on translation and repetition.Khazaie and Ketabi (2011) propose some types of vocabulary learning materials based on different LCR with or without pictorial or written annotations.Their outcomes demonstrate that learning materials with written or pictorial annotations help improve the performance of students.In addition, Mashhadi and Jamalifar (2015) analyze the effect on intermediate learners' vocabulary learning using textual and visual cues.Their results claim that learners that learn new vocabulary by pictorial and textual cues improve their vocabulary knowledge and rarely face difficulty in remembering them.In this way, Table 1 summarizes some LCR types and the information that can be displayed by each channel to teach new vocabulary via mobiles.
As can be seen in Table 1, the main difference among the three proposed LCR types is the kind of information provided by each channel (see Figure 1).In the case of the visual channel, the app can show either an image related to the meaning of a word, or the word (see Figure 1a).In the case of the auditory modality, learners can listen to the pronunciation of a word, or can listen to an audio related to the meaning of the word (see Figure 1b).The tactile channel defines how the learners interact with the interface, for example, they can tap on a button to continue with the next word, or they can tap on a specific image to obtain information related to the word (see Figure 1c).Therefore, the use of an adequate learning strategy in combination with the correct LCR type could improve the acquisition of new vocabulary, but few experiments have been carried out to discover the most appropriate LCR type to learn and memorize new words using m-learning.

Current Study
Zimmerman ( 2001) defines a self-regulated learning strategy (SRL) as the ability of students to become masters of their own learning processes.A self-regulated learning approach promotes the students' self-initiated processes for improving their methods of learning.Seker Learners listen to the pronunciation of a word.
Learners tap on a button to continue with the next word.
Option 2 Learners see the word.
Learners listen to audio related to the meaning of the word.
Learners tap on a button to continue with the next word.Option 3 Learners see a list of words.
Learners listen to the pronunciation of each word Learners tap on a specific word and then they see one image related to the meaning of the word and listen to the pronunciation.
(2016) claims that incorporating SRL strategies to teach foreign languages improve the development of autonomous learners, but professors mostly do not consider the use of these strategies to teach.A SRL approach considers learning as a task that students do for themselves in an active way; in other words, the learning is not constrained to a specific process, a specific order or learning technique.
On the other hand, multimodal interaction, provided by mobile devices or tablets, can be used to promote either a self-regulated strategy or a non-self-regulated learning strategy.The implementation of one or another learning strategy depends on the LCR type used and the interaction with the application.Previous studies (Al-Samarraie et al., 2016; N. S. Chen et al., 2008;C. J. Chen & Liu, 2011) demonstrate that LCR types could affect the students' learning process.Consequently, the next research question was addressed: do the use of an LCR type based on a self-regulated learning strategy improve the vocabulary learning performance of students using m-learning?To address this question, this study investigates the relationship between two LCR types and the students' academic performance to memorize new vocabulary.For this study, the following LCR types were designed and implemented: 1.The NSRL type (see Figure 2a): this LCR type promotes a non-self-regulated learning strategy.
It provides information about a specific word at a time displaying an image related to the meaning of the word and the words in the native language and the foreign language.In this type, the order to memorize the vocabulary is fixed, so students must tap on the ''next'' button to continue with the next word on the list.2. The SRL type (see Figure 2b): this LCR type promotes a self-regulated learning strategy.It displays a set of images related to the meaning of different words.In this type, the order to memorize the vocabulary is not fixed, so students must tap each image on the screen to show the word in the foreign language.As can be seen in Figure 2, both LCR types use the visual channel to display an image related to the meaning of a word and the word, and the auditory channel is used to play a sound with the pronunciation of the word.The main difference between both LCR types is the learning strategy.In the case of the NSRL type, students must tap on the ''next'' button to continue with the next word in the sequence.When the sequence is finished, students can start again, but they are not allowed to select and rehearse a specific word.In this case, the sequence of the words is fixed to promote a non-self-regulated learning strategy.On the other hand, in the case of the SRL type, students can tap any image on the screen to know its translation, so they are free to memorize the words in whichever order they choose.In this case, students can select and rehearse one specific word as many times as necessary to learn it and to promote a self-regulated learning strategy.

Experimental Task
The experiment was run in the facilities of the School of Engineering in a public Mexican University.All participants were native speakers of Spanish, and as part of their study program they received English courses.The experimental task selected for this study was to learn 10 words in German using m-learning.In terms of difficulty, German is considered to be a Category II language that increases it in difficulty, and it is harder to learn than another Category I language like Italian, French or Portuguese.Each student had 15 min to memorize the list of words, and then answer an exam to evaluate their knowledge.The ten German words were the same for both experimental groups (10 farm animals, see Figure 2).

Hardware and Software Set-Up
For this study, we used a quad-core 1.20 GHz Cortex-A9 tablet with 1 Gb of RAM and 10.1 in of a capacitive touchscreen with a resolution of 1,280 3 800 pixels (without an Internet connection to avoid distractions).The tablet was used by the participants to perform the training sessions.The test application was developed in App Inventor for Android 4.2 (Wolber, 2014).During the test, participants were seated, and the evaluator was only responsible for observing their performance and to give instructions at the beginning and end of the test.Figure 3 shows the experimental setup and one of the participants who performed the test interacting with the tablet and answered an exam.

Participants
Forty undergraduate students participated in this study (mean age = 19.85 6 2.19).None of them knew the German language.They were divided into two experimental groups: NSRL Group: 20 students were trained using the NSRL strategy.SRL Group: 20 students were trained using the SRL strategy.
All students reported being healthy, a normal sense of sight, hearing, and touch, and none of them reported a history of psychiatric or neurological pathology.Besides, each participant was informed about the objective and procedure of the experiment before starting the test.Informed consent was obtained from all individual participants included in the study.All procedures performed in studies involving human participants were in accordance with the 1964 Declaration of Helsinki.

Instruments and Data Collection
The student's capacity for memorization was assessed using the meta-memory subset of the Neuropsychological Battery of Executive Functions and Frontal Lobes Version-2 test (known as BANFE in Spanish; (Flores Laa¨zaro et al., 2017).This test has been developed and standardized for Spanish-speaking populations.The test helps to avoid bias due to the difference in the ability of the participants to memorize and remember words.This subtest has five trials in which participants have to predict how many of nine words they could learn and remember in each trial.The results of this test were used to perform the assignment of participants to each experimental group (pseudorandom process).The evaluator calculated the normalized negative errors (NNE) and the normalized positive errors (NPE) of each participant for further analysis.
To evaluate the students' performance in each experimental group, data were collected during the training session of each participant.In the case of the NSRL Group, the app automatically recorded the number of times that the sequence was repeated, and the number of studied elements in the current sequence (when the sequence was not finished).For the SRL Group, the app recorded the number of times that each figure on the screen was tapped.
Usually, teachers have trouble evaluating students' vocabulary acquisition.According to Gathercole and Adams (1994) and Kaushanskaya et al. (2011), shortterm memory is an important predictor of vocabulary acquisition.One of the most widely used methods to assess vocabulary acquisition is to apply a written test to measure the number of words memorized (Jia et al., 2012).In this study, the academic performance of each student was measured using the grade of a written exam that had a perfect score of 10 points (the exam was administered on paper, see Figure 3b), in which the evaluator gave the students Spanish words to write in German (to get each point the word had to be written correctly without any error).

Intervention
Each participant was scheduled for two sessions over 2 days.On the first day, the participant was interviewed and the BANFE test was applied.On the second day, the test started with a familiarization phase, continuing with the training session (to memorize the new vocabulary), and ended with the evaluation test.The overall procedure of the test is presented in Figure 4.Each stage is described below: 1. Interview: at the beginning, the evaluator explained the objective of the test and the experimental task.Students who agreed to continue filled out an identification sheet and signed a consent form.2. BANFE test: in this stage, the metamemory subset of the BANFE test was used to evaluate the student's ability to memorize and remember words.This test ensured the homogeneity of both experimental groups which allowed us to assign the participants to each experimental group and reduce bias in the results due to the subjects' ability to recall words.(Bacon & Izaute, 2014).3. Familiarization: the students were instructed to use the m-learning app.They also received a brief demonstration of how the corresponding LCR type worked.When students performed this stage, to avoid the learning effect they used a different task than the one used in the training session (in this case, they had to learn the name of three colors in German).The objective of this step was to make sure that students did not have any doubts about how they should interact with the app.4. Training: the students had 15 min to memorize 10 words in German.Participants from the NSRL group, could repeat the sequence as many times as necessary to learn the words.On the other hand, participants from the SRL group, could select and rehearse a specific word as many times as necessary to learn it.After 15 min, the application closed by itself.During the training, the app automatically recorded the students' interaction with the app.The main difference between both experimental groups is the type of intervention that receive the subject in each group.In this way, the performance of each student (ability to recall words) will depend on the strategy they received to memorize the words.5. Evaluation test: the students in both groups were asked to complete an exam to evaluate their performance in memorizing the new vocabulary.The content of the exam consisted of reading 10 Spanish words and writing the correct German words.

Data Analysis
The software Minitab (ver.18,0; Minitab Inc, USA) was used to perform descriptive statistical analysis, a twosample t-test, and boxplots with the numeric data of the training sessions and evaluation tests of each participant.
All participants finished the experimental procedure, so no data were discarded.The ability of participants in each experimental group to memorize words were analyzed using the normalized negative errors (NNE) and the normalized positive errors (NPE) of the metamemory subset of the BANFE test.

Results
As mentioned in the previous section, participants were pseudo randomly assigned to each experimental group.The assignment was based on the score of the metamemory subset of the BANFE test.With the purpose of verifying that there was no bias in the results due to the ability to memorize the participants in each group, the NNE and NNP scores of each group were statistically compared.
A two-sample t-test was performed to verify that there was not statistically significant difference between the two experimental groups.Two-sample t-test was used because although the participants were pseudo randomly assigned to each experimental group the data of the two samples are statistically independent.
To use the two-sample t-test, it is necessary to test that the data from both groups are normally distributed.One of the primary reasons for employing a test for normality is that the t-test assumes that the data comes from a normal distribution.If this assumption is violated, the reliability and validity of the t-test results can be compromised.
The Ryan-Joiner Normality (RJ) Test, a derivative of the Shapiro-Wilk test, is widely recognized for its efficiency and power, especially for small to moderate sample sizes (Minitab, 2023).Unlike many other tests of normality, the RJ test has a high sensitivity to detect departures from normality, which means it's less likely to indicate that data is normal when it's not (a Type II error).This is crucial in research scenarios, as overlooking non-normality can lead to incorrect statistical conclusions (Chantarson, 2015;Seir, 2011).
Ryan-Joiner's method calculates the correlation between the data and the normal scores of the data; values of RJ closer to 1 indicate that the data are normally distributed.Another advantage of this test is its interpretability; a value closer to 1 can be easily understood as indicating stronger conformity to normal distribution, allowing for intuitive understanding even for those not deeply versed in statistical methods.
The normality test showed that data followed a normal distribution (NNE, RJ = 0.98, and NPE, RJ = 0.93).The values obtained in this study affirm the normality of the distribution and hence justify the use of the two-sample t-test for further analysis.
A two-sample t-test with a significance level of a = .05was performed to compare NNE score between NSRL and SRL groups; t(df) =20.29, p = .77;and to compare NPE score between NSRL and SRL groups; t(df) =20.83, p = .41.
The descriptive statistical analysis of the students' performance is shown in Table 2, and its boxplot is shown in Figure 5.As can be seen, the mean of the student's performance of the SRL group was higher than the student's performance of the NSRL group.In both groups, a Ryan-Joiner Normality Test was run, and the data followed a normal distribution (SRL, RJ = 0.96, and NSRL, RJ = 0.98), consequently, a two-sample t-test (a = .05)demonstrated that there was a significant difference (p = .002)and Cohen's d = 1.07 with a large effect size between both experimental groups, it means that the difference in subject's ability to memorize words can be attributed to the experimental intervention.
For 15 min, the participants were able to train and memorize the 10 words in German.The participants in the NSRL group trained each word one by one using a non-self-regulated strategy, in other words, the order to memorize the words was fixed.When the sequence was finished, participants could start again from the first word.As it was mentioned before, the app automatically recorded the number of times that the sequence was repeated.On the other hand, the participants from the SRL group trained each word without any restriction using a self-regulated strategy.They did not have to follow a fixed order to memorize the words and they were able to tap any image on the screen as many times as necessary to memorize each word.In this case, the app recorded the number of times that each figure was tapped.
In the case of the NSRL group, the Number of Trained Words (NTW NSRL ) of each participant was calculated using Equation 1(where: a = number of times that the sequence was repeated, and b= number of words in the current sequence).
In the case of the SRL group, the Number of Trained Words (NTW SRL ) of each participant was calculated using Equation 2(where: n1 = number of times that the figure 1 was tapped on, n2 = number of times that the figure 2 was tapped on, n3 = number of times that the figure 3 was tapped on, etc.).
Figure 6 shows a boxplot to compare the total number of elements studied by students of each experimental group.It shows that the mean of the total number of elements studied in the SRL group is higher in comparison with the value of the NSRL group.In addition, a twosample t-test (a = .05)demonstrates that there is a significant difference (p = .001)between both experimental groups.
Figure 7 shows the scatterplots to compare the total number of repetitions with the students' performance in each experimental group.A two-sample t-test (a = .05)demonstrates that there is no correlation between the number of repetitions of each word and the performance of students from the NSRL (p = .16)and SRL groups (p = .43).

Discussion
This study addresses the following question: do the use of an LCR type based on a self-regulated learning strategy improve the vocabulary learning performance of students using m-learning?The results demonstrate that there is a relationship between the LCR type and the students' performance because students learning with the SRL strategy outperform those learning with the NSRL strategy (Figure 5).In addition, the statistical tests show that there is a significant difference between the two proposed LCR types (p = .002,a = .05).This fact means that the use of LCR types that promote a self-regulated learning strategy improves the ability of the students to memorize words in a foreign language.These results reinforce other studies that show that technology-assisted incidental vocabulary learning aims to help learners acquire words incidentally through digital technologies (Yu & Trainin, 2022).One of the main aspects that this study contributes consists in determining the effectiveness of the different representation systems for learning the vocabulary of a second language: written, visual and auditory from the principles of ubiquitous and mobile learning that allow constant feedback in the vocabulary acquisition.Research on vocabulary learning through apps is still scarce and the main studies have focused on the ''Quizlet'' application.This type of application has the limitation of only integrating three modes of representation: writing, spelling, and matching activities; not allowing to develop and compare the sounds of two words and their translation in different languages.What we have shown in this study is that a more complete representation that includes the translation and pronunciation of the word can favor a more effective vocabulary learning.In this sense, the scientific literature has shown that when the representation systems are diverse they allow a more active and long-lasting learning of vocabulary (Dreyer, 2014;Kingsley & Grabner-Hagen, 2018).When this learning is also carried out from a gamified approach, the effects on vocabulary learning increase (Va´zquez-Cano et al., 2023;Weissheimer et al., 2019).
Additionally, the metamemory subset of the BANFE test shows that there is not statistically significant difference between the two experimental groups concerning participants' ability to memorize words.This fact shows that in this study the student's academic performance is influenced only by the LCR type used in m-learning.According to the results in Figure 6, there is a significant difference (p = .001)between the number of studied elements in both experimental groups.It means that participants from the SRL group repeated the words more times in comparison with the participants from the NSRL group.
This study provides another fundamental result that contradicts much of the published literature on vocabulary acquisition.The results show that repetition is not a key element in the effectiveness of vocabulary learning when there are multiple representation formats.The students with the highest scores (grades) in both groups are not the ones with the highest total number of repetitions.This fact means that the learning of new vocabulary does not depend on the number of repetitions.This result is in contrast with previous studies (English & Visser, 2014;Kuhl & Anderson, 2011) that mention that memory improves with repetition.Moreover, according to Figure 7 most of the students that learn new vocabulary using the SRL type could have a grade higher than 6.0, while using the NSRL type most of them could have a grade lower than 6.0, but new experiments should be run to obtain conclusive results.
Regarding the relationship between the two proposed LCR types, the SRL type promotes a self-regulated strategy in which, according to Zimmerman (2001), students become masters of their own learning processes and improve their methods of learning.From the observations of the evaluator, the strategy to memorize words used by each participant from the SRL group was not the same.For example, some students started tapping on one figure and repeated the same figure several times until they learned the word before tapping on the next one.Other students started tapping on one figure twice or three times and changed to the next one, and later they repeated the previous figure to verify that they had learned the word.According to Garcia (1996), a selfregulated strategy allows students to rely on their own resources, proactively seek information, and autonomously implement selected strategies or skills in order to achieve the learning goals.In this sense, the LCR-SRL type could reduce students' cognitive load, enhancing the capacity limit associated with short-term memory leading to higher achievement (Cowan, 2011;Paas & Van Merrie¨nboer, 1994).In contrast, the NSRL type limits students' ability to promote their own learning strategies because the learning content is presented in a fixed order.
Furthermore, incorporating different representations of learning content into second language learning is consistent with established cognitive theory and teaching practice (Chang et al., 2022;C.-M. Chen, 2009).It enriches the learning experience by enhancing cognitive processing, providing context, using multiple sensory modalities, accommodating individual learning styles, and promoting motivation and engagement (Mathias et al., 2022).Together, these factors contribute to more effective second language learning.
In this sense, when different typologies of images are combined when associating visual representations with sounds, the cognitive load in second language vocabulary learning is increased.These strategies have been considered as one of the best didactic strategies for contextualized second language vocabulary learning (Butler et al., 2008;Cook & Gor, 2015).The combination of multimodal approaches enhances vocabulary consolidation and retention as well as improves phonological awareness and promotes cross-modal integration (Janssen et al., 2015;Kormos & Sa´fa´r, 2008).Also, learning content representations can be used repetitively to promote feedback in vocabulary learning; therefore, these didactic strategies based on multimodal mobile learning are particularly interesting as they promote ubiquitous learning without parental or teacher control, increasing selfregulated learning (Bannert et al., 2009;Deunk et al., 2018;Sa´ez-Lo´pez et al., 2023).When learners associate words with visual representations, they leverage this powerful memory system to enhance vocabulary retention.The integration of visual and auditory information can stimulate the brain's hippocampus, a region critical for memory consolidation (Ibarra-Santacruz & Martı´nez-Ortega, 2018).Furthermore, engaging multiple sensory modalities with images and sounds can make the learning experience more enjoyable and motivating (Mathias et al., 2022).
Some authors have explored self-regulated strategies for vocabulary learning.Ping and Siraj (Ping & Siraj, 2012) investigated the use of self-regulated learning strategies for vocabulary learning.Their results showed that low processing vocabulary learning strategies (i.e., oral repetition and using dictionary for comprehension purpose) are dominant among the students, while deep processing strategies (i.e., semantic grouping and word structure) are less used by the learners.Ping et al. (Ping et al., 2015) reported a preliminary study for developing strategy instruction for Chinese English Foreign language learners.Their findings suggested that there is a pressing need to enhance self-regulated learning strategies in vocabulary learning because these strategies are rarely applied by the learners.In addition, these LCR-SRL strategies can favor greater effectiveness when the linguistic distance (degree of closeness between languages) is low (Spanish-German) (Chiswick & Miller, 2012).Furthermore, the multiple representation design presented in this study could help to foster greater intrinsic motivation and an increase in student autonomy (Almusharraf, 2020;Nii & Yunus, 2022).This type of apps also allows a combined use between the classroom and the house; which favors greater feedback and support for students with learning difficulties.At the same time, it introduces a playful element that increases motivation (Chik, 2014;Sato et al., 2013); although this motivation must be combined intrinsically and extrinsically through teacher and parental support to ensure that learning and motivation do not decrease (Buckley & Doyle, 2016).
The pedagogical implication of this study is that LCR-SRL types must be used in m-learning to improve students' performance for vocabulary learning.Future studies should focus on how to integrate low processing and deep processing strategies in m-learning and understand its effects on improving student's vocabulary acquisition.

Conclusions
Some authors have demonstrated that mobile devices are effective, efficient, and engaging for learning, and define the term m-learning as the use of mobile technology as an education tool to access learning content anywhere and anytime.Using mobile devices, users can access learning content through the visual, auditory, and tactile channels.Moreover, the use of m-learning to teach a second language contributes to the development of four language skills (listening, reading, writing, and speaking).Nevertheless, some authors have reported negative effects on the use of m-learning and more studies should be run to provide a conclusion about how new vocabulary is better acquired using mobile technology.
The current study was focused on understanding the effects on improving the students' academic performance using two LCR types in a vocabulary teaching task.An experiment with 40 undergraduate engineering students was run to analyze the relationship between the LCR types and the students' academic performance to learn new vocabulary.The results claim that there is a relationship between the LCR types and the students' performance.Students who learn using a self-regulatedbased LCR type perform better than students who learn using a non-self-regulated-based LCR type.This fact is significant in that it provides a signal that teachers must consider the use of applications based on SRL types for vocabulary acquisition tasks.Moreover, developers must consider during the design of an m-learning application the use of LCR types that promotes self-regulated learning strategies.
The strength of this study lies in that the participants of each experimental group were assigned according to his or her ability to memorize words in order to avoid the bias in the results due to this fact.The SRL type seems to be effective in improving the students' learning abilities because, from the observations of the evaluator, participants in the SRL group did not use the same strategy to memorize the new words, the authors suspect that these students maybe use more complex memorization techniques.
The experimental results suggest some directions for future works.The outcomes show that participants from the SRL group repeated the words more times in comparison with the participants from the NSRL group.Nevertheless, the statistical analysis shows that there is not a significant correlation between the number of repetitions of each word and the students' performance.This fact is in contrast with the results of English and Visser (2014) and Kuhl and Anderson (2011), that claim that memory improves with repetition.However, the design of the experimental task was focused on analyzing the relationship between the two LCR types and the student's performance, the number of repetitions of each word was not considered.In this way, the authors propose to run new experiments to analyze the relationship among the LCR types, the number of repetitions of the words, and the effects on the student's performance.
On the other hand, the results of this study are valid to describe the students' ability to memorize new words in class (only short-term memory was measured); the authors suggest that new experiments should be run to analyze if the results are valid for long-term memory to add reliability and generalizability of our findings.

Figure 1 .
Figure 1.Examples of visual, auditory and tactile modalities to provide information on the word cow in German (kuh): (a) visual channel, (b) auditory channel and c) tactile channel.

Figure 2 .
Figure 2. (a) NSRL and (b) SRL types for learning German vocabulary using m-learning.

Figure 3 .
Figure 3. Experimental set-up: (a) one student is learning new vocabulary using the app and (b) one student is answering the exam to evaluate his performance.

Figure 5 .
Figure 5. Boxplot of the students' performance of both experimental groups.

Figure 6 .
Figure 6.Boxplot of the total number of trained words in each group.

Figure 7 .
Figure 7. Scatterplots to compare the total number of repetitions and the student's performance in each experimental group.
for the scholarship CVU 631308 to carry out this research.This research has been developed with the support of the I + D + I Project: ''Proof of Concept'' Project entitled: ''GAUBI-ORTO.Evaluation of a digital model and application for sustainable, ubiquitous and gamified learning of Spanish spelling in Primary Education'' (PDC2022-133185-I00). State Program to Promote Scientific-Technical search and its Transfer, of the State Plan for Scientific and Technical Research and Innovation.Ministry of Science and Innovation (Spain).

Table 1 .
Examples of LCR Types That Can be Used for Teaching New Vocabulary Using m-learning.

Table 2 .
Statistical Analysis of Both Experimental Groups.