Mobile learning technologies

MLTs refer to the use of mobile devices and technologies to facilitate and enhance learning experiences [25], with growing recognition of their potential to provide flexible, accessible, and personalized learning opportunities [26]. They have been increasingly employed in music education to support various aspects of musical learning [27–32], with evidence showing effectiveness in enhancing learning outcomes such as improved academic performance, increased engagement, and higher satisfaction [33–35].

The use of MLTs in music education, particularly in piano learning, can be categorized into three main types: mobile apps, virtual tools, and learning platforms. Mobile apps for piano learning offer a range of features and functionalities, such as interactive lessons, real-time feedback, and gamification elements [36]. These apps provide benefits such as increased accessibility, flexibility, and engagement in piano learning [37]. Popular mobile apps for piano learning include Simply Piano, Piano Maestro, and Yousician [38]. Virtual tools, such as virtual keyboards and virtual piano tutors, offer advantages like enhanced visualization, immediate feedback, and the ability to practice without access to a physical piano [39,40]. Learning platforms for piano learning are characterized by comprehensive curricula, structured lessons, and progress tracking features [38]. These platforms offer benefits such as personalized learning paths, interactive content, and social learning opportunities [41].

MLTs have the potential to transform piano learning by enhancing accessibility and flexibility, allowing learners to access resources and practice at their convenience [26]. MLTs can also facilitate self-regulated learning by providing learners with tools to set goals, monitor progress, and self-evaluate their performance [42]. Furthermore, MLTs can offer personalized learning experiences by adapting content and feedback based on individual learners’ needs and preferences [34]. Finally, MLTs can foster learner engagement and motivation through interactive features, gamification, and social learning opportunities [43].

Hypothesis development

Several factors have been identified as potential antecedents to the adoption and use of MLTs in piano learning. This study considers three key factors: self-efficacy in the use of MLTs, prior domain-specific knowledge, and prior experience with technology-mediated learning. These factors were selected based on their theoretical relevance and empirical evidence from previous studies in the fields of technology adoption, music education, and mobile learning [23,44,45].

Self-efficacy is a key concept in social cognitive theory, which refers to an individual’s belief in their ability to successfully perform specific tasks or achieve desired outcomes. It is a context-specific construct that influences an individual’s thoughts, feelings, and behaviors, shaping their motivation, perseverance, and resilience in the face of challenges [46]. Self-efficacy beliefs are formed through various sources, including mastery experiences, vicarious experiences, verbal persuasion, and physiological and emotional states [47].

Self-efficacy in the use of MLTs refers to an individual’s belief in their ability to successfully utilize mobile devices and applications to support their learning processes. Self-efficacy encompasses a learner’s confidence in their skills to navigate, interact with, and effectively use mobile apps, virtual tools, and learning platforms to enhance their piano playing abilities [45]. Self-efficacy is a crucial factor to consider in the adoption and use of MLTs for piano learning because it influences learners’ perceptions of the technologies’ usefulness and ease of use, their willingness to engage with the learning content, and their persistence in the face of challenges [23,24]. This relationship aligns with established technology acceptance models, where self-efficacy serves as a key determinant of behavioral intention and actual technology use, as individuals with higher self-efficacy are more likely to perceive technology as beneficial and less threatening, thereby facilitating acceptance and adoption [22]. Learners with higher levels of self-efficacy are more likely to perceive these technologies as valuable tools for enhancing their learning experience. They are more likely to explore the features and functionalities of MLTs, engage with the learning content, and persist in the face of challenges [45]. Conversely, learners with lower levels of self-efficacy may be more hesitant to adopt MLTs, as they may perceive these technologies as difficult to use or doubt their ability to benefit from them [44]. Therefore, the first hypothesis of this study is as follows:

H1: Self-efficacy in using MLTs positively affects MLT adoption in piano learning.

Prior domain-specific knowledge refers to a learner’s pre-existing knowledge and skills related to piano playing, such as music theory, technique, and repertoire [48]. It represents the foundation upon which learners can build and expand their piano playing abilities. Learners with higher levels of prior domain-specific knowledge are better equipped to understand and apply the concepts, strategies, and feedback provided by MLTs [45]. Moreover, prior domain-specific knowledge can help learners to recognize the value and relevance of MLTs in supporting their piano learning goals. Learners who have a solid understanding of music theory and piano technique are more likely to appreciate how MLTs can help them reinforce their knowledge, refine their skills, and explore new repertoire [44]. In contrast, learners with limited prior domain-specific knowledge may find it more challenging to navigate and benefit from MLTs, as they may lack the necessary foundation to fully engage with the learning content [45]. Hence, the current study formulates the second hypothesis as follows:

H2: Prior domain-specific knowledge positively affects MLT adoption in piano learning.

Prior experience with technology-mediated learning refers to a learner’s previous exposure to and engagement with educational technologies and online learning environments [23]. It encompasses a learner’s familiarity with using digital tools, navigating online platforms, and engaging in self-regulated learning in technology-mediated contexts. Learners who have positive experiences with technology-mediated learning are more likely to perceive MLTs as useful and easy to use, leading to increased adoption and use [45].

Prior experience with technology-mediated learning can also help learners to develop the necessary skills and strategies to effectively use MLTs for piano learning. Learners who have successfully navigated other educational technologies may be more adept at exploring the features and functionalities of MLTs, setting learning goals, and monitoring their progress [44]. They may also be more comfortable with the self-directed nature of learning with MLTs, as they have developed the metacognitive strategies necessary for effective self-regulated learning [45]. In contrast, learners with limited prior experience with technology-mediated learning may face a steeper learning curve when adopting MLTs for piano learning. They may require more support and guidance in navigating the platforms, understanding the learning content, and developing effective learning strategies [23]. As such, prior experience with technology-mediated learning can play a significant role in shaping learners’ attitudes towards and use of MLTs for piano learning. Thus, the third hypothesis of this study is written as follows:

H3: Prior experience with technology-mediated learning positively affects MLT adoption in piano learning.

MLTs have been increasingly recognized for their potential to enhance learning outcomes across various educational domains, including music education and piano learning. Numerous studies have investigated the impact of MLTs on learners’ performance, engagement, and motivation, providing evidence for the effectiveness of these technologies in supporting learning processes [26,35]. For instance, mobile apps and virtual tools that offer interactive content, real-time feedback, and gamification elements have been found to improve learners’ technical skills, musical expression, and creativity in piano playing [36,37]. Similarly, learning platforms that provide comprehensive curricula, personalized learning paths, and progress tracking features have been shown to enhance learners’ engagement, motivation, and self-regulated learning in piano education [38,41].

The positive impact of MLTs on learning outcomes can be attributed to several factors. First, MLTs provide learners with flexible and accessible learning opportunities, allowing them to engage with educational content at their own pace and convenience [42]. This flexibility is particularly beneficial for piano learners, who can practice and develop their skills whenever and wherever they have access to a mobile device [49]. Second, MLTs offer personalized learning experiences tailored to individual learners’ needs, preferences, and skill levels [34]. Adaptive learning algorithms and intelligent tutoring systems embedded in MLTs can provide learners with customized feedback, recommendations, and challenges, optimizing their learning outcomes [41]. Finally, MLTs can foster learner engagement and motivation through interactive features, gamification, and social learning opportunities [43]. By providing learners with engaging content, rewards, and a sense of community, MLTs can create a more enjoyable and rewarding piano learning experience, ultimately leading to improved learning outcomes [41].

H4: The use of MLTs positively affects learning outcomes in piano education.

Based on the hypotheses presented in the attached document, the proposed theoretical framework for this study investigates the factors influencing the adoption and use of MLTs in piano learning and their impact on learning outcomes. As depicted in Fig 1, the framework considers three key antecedents: self-efficacy in the use of MLTs, prior domain-specific knowledge, and prior experience with technology-mediated learning. These antecedents are hypothesized to positively influence the adoption and use of MLTs for piano learning (H1, H2, and H3). In turn, the use of MLTs is hypothesized to positively influence learning outcomes in piano education, including technical proficiency, musical expression and creativity, and engagement and motivation (H4). This framework aims to provide a comprehensive understanding of the factors that contribute to the successful adoption and use of MLTs in piano learning and their potential benefits for learners’ performance and experience. Bringing these constructs together, we propose the following research model.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Proposed PLS-SEM model of mobile learning technology adoption.

https://doi.org/10.1371/journal.pone.0330269.g001

Data analysis

To test the proposed conceptual framework and the associated hypotheses, this study employs Partial Least Squares Structural Equation Modeling (PLS-SEM) using SmartPLS software. SEM is a multivariate statistical technique that allows for the simultaneous analysis of multiple relationships between observed and latent variables [50]. SEM is considered superior to traditional regression techniques because it enables researchers to test complex models with multiple dependent and independent variables, while accounting for measurement error [51].

PLS-SEM is a variance-based approach to SEM that focuses on maximizing the explained variance of the dependent latent variables [50]. PLS-SEM is particularly suitable for this study because it is effective in handling complex models with a large number of constructs and relationships, even with relatively small sample sizes [52]. Moreover, PLS-SEM is more robust to deviations from normality compared to covariance-based SEM techniques [50]. SmartPLS [53] is a user-friendly software that enables researchers to create and estimate PLS-SEM models. It provides a graphical user interface for model specification and offers various evaluation criteria for assessing the measurement and structural models [50].

In PLS-SEM, the model consists of two main components: the measurement model and the structural model. The measurement model specifies the relationships between the observed variables (indicators) and their corresponding latent variables (constructs). The structural model represents the hypothesized relationships between the latent variables [50]. Evaluation of the measurement model involves assessing the reliability and validity of the constructs. Reliability is examined using Cronbach’s alpha and composite reliability, with values above 0.7 indicating good reliability [50]. Convergent validity is assessed by examining the average variance extracted (AVE), with values above 0.5 indicating adequate convergence [54]. Discriminant validity is evaluated using the Fornell-Larcker criterion and cross-loadings, ensuring that each construct is distinct from the others [50].

The structural model is evaluated by examining the path coefficients, their significance (using bootstrapping with 5,000 resamples and 95% confidence intervals), and the coefficients of determination (R² values) for the endogenous latent variables. Path coefficients in SEM represent the direct effects of predictor variables on outcome variables and quantify the strength and direction of relationships between constructs, typically ranging from −1 to +1. They are crucial for interpreting results, with significant p-values (typically <0.05) indicating support for hypothesized relationships and values closer to ±1 suggesting stronger influences [50]. The R² values measure the predictive power of the model, with higher values indicating better explanatory power [55]. The goodness-of-fit (GoF) index, although not a primary evaluation criterion for PLS-SEM, can be used to assess the overall fit of the model, with values ranging from 0 to 1, where higher values indicate better fit [56].

Data collection

This study employed a cross-sectional design, focusing on two distinct samples: university students studying music in China and Russia. The selection of universities was based on a purposive sampling approach, considering factors such as program reputation, geographical diversity, and accessibility. In China, a top-tier conservatory located in Beijing was chosen, known for its comprehensive music programs and national recognition. In Russia, a prestigious music academy in Moscow was selected, representing one of the country’s leading institutions for musical education. While these institutions are prominent in their respective countries, it is important to note that they may not be fully representative of all music education programs in China and Russia.

The names of the universities were kept anonymous to protect the privacy and confidentiality of the participants and to adhere to the ethical guidelines of the study. All the data were gathered anonymously during the period of research. Written consent was obtained from the participants.

The target population consisted of bachelor’s level music students enrolled in these selected universities. To ensure the comparability and homogeneity of the samples, strict inclusion criteria were established. Participants were required to be (1) full-time bachelor’s students, (2) majoring in music, (3) enrolled in the selected universities during the data collection period, and (4) willing to voluntarily participate in the study. Students who did not meet these criteria were excluded from the study.

A stratified random sampling method was utilized to select participants from each university. This sampling approach ensures that the sample is representative of the population within each institution and reduces potential sampling bias [57]. First, the music department in each university was identified as the primary stratum. Then, within each department, students were further stratified based on their year of study (i.e., first, second, third, or fourth year). From each stratum, a random sample of 30 students was selected using a random number generator, resulting in a total sample size of 120 students per university. This sample size is considered adequate for PLS-SEM analysis, as it exceeds the recommended minimum of 10 times the number of paths directed at any latent variable in the model [50].

Ethical approval

The research project has undergone thorough review by the Office of Research Ethics Committee at Baoshan University. Following a comprehensive evaluation, the committee determined that the study qualifies for exemption from full ethical review based on the following criteria: 1) the research adheres to the principles outlined in the 1964 Helsinki Declaration and its subsequent amendments, 2) all interactions with participants maintain strict anonymity throughout the study, 3) informed consent procedures were appropriately implemented and documented, 4) the nature of the study does not require additional ethical oversight beyond standard research practices.Informed consent was obtained from all subjects involved in the study or their legal guardians.

To ensure ethical conduct throughout the study, participants were provided with informed consent forms that clearly outlined the purpose of the research, assured confidentiality and anonymity, and emphasized the voluntary nature of their participation, allowing them to withdraw at any time without consequences. The data collection process was conducted in close collaboration with the music departments of each university. A questionnaire using 5-point Likert-like scale was developed where 1 means strongly disagree, 2 means disagree, 3 means neutral, 4 means agree, and 5 means strongly agree (see Table 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Evaluation questions of the main variables of the current study.

https://doi.org/10.1371/journal.pone.0330269.t001

The questionnaire items were developed through a comprehensive review of existing literature on mobile learning technologies, music education, and technology adoption in educational contexts. The instrument employed a self-report methodology, allowing participants to provide subjective assessments of their experiences and perceptions across all constructs. For the Learning Outcomes construct, participants were asked to self-evaluate their perceived improvements in piano skills, practice frequency, lesson difficulty, and confidence levels after engaging with online piano learning resources, as self-evaluation methods capture learners’ subjective experiences and perceived effectiveness of educational interventions, which are often more predictive of continued engagement than objective performance measures [58]. To measure Prior Domain-Specific Knowledge, years of formal piano lessons and practice frequency were employed, as years of experience represents a common approach to assess prior knowledge across different domains [59], and these temporal measures reflect stable, consolidated knowledge that differs from recently acquired background information [60]. Items measuring Prior Experience with Technology-Mediated Learning were developed by assessing participants’ frequency of mobile device usage for learning, comfort with technology-mediated learning, and perceived helpfulness of these technologies, as prior experience with educational technology has been identified as a key predictor of technology acceptance and successful learning outcomes [22].

Items measuring Self-efficacy in the use of MLTs were developed by assessing participants’ confidence in navigating mobile learning technologies, their perceived skills in effectively utilizing these tools, and their comfort in exploring new features, as self-efficacy beliefs are fundamental predictors of technology acceptance and usage behavior [47,61]. Finally, Items assessing Mobile Learning Technology Adoption were formulated to capture participants’ regular usage of mobile apps, virtual tools, and online learning platforms for piano learning, reflecting established patterns in technology adoption measurement that focus on actual usage behaviors and user experiences. Prior to data collection, the questionnaire underwent expert validation by three music education specialists and two technology adoption researchers to ensure content validity and cultural appropriateness. A pilot test was conducted with 30 music students (15 Chinese, 15 Russian) to assess item clarity and reliability. Minor wording adjustments were made based on pilot feedback, and Cronbach’s alpha values exceeded 0.70 for all constructs in the pilot sample, confirming the instrument’s reliability. This questionnaire was administered online using Qualtrics (a secure survey platform), which included informed consent, demographic questions, and measures for each construct in the proposed conceptual model. The survey was translated into the native languages of the participants (Chinese and Russian) and back-translated to ensure accuracy and equivalence by the authors of the current article, who are fluent in English and Chinese, and English and Russian, respectively [62].

Invitation emails containing a link to the online survey were sent to the randomly selected students’ university email addresses. The email emphasized the voluntary nature of participation, assured confidentiality, and provided an estimate of the time required to complete the survey (approximately 15–20 minutes). Two reminder emails were sent at one-week intervals to encourage participation and maximize the response rate [63]. The data collection process was conducted in August 2023.

Evaluation of measurement model

Table 2 presents the reliability and validity measures for the variables used in the study, including Cronbach’s Alpha, Composite Reliability, and Average Variance Extracted (AVE). Cronbach’s Alpha is a measure of internal consistency reliability, assessing how well the items within each variable are related to one another. It ranges from 0 to 1, with higher values indicating better reliability. The Cronbach’s Alpha values for the variables in this study range from 0.77 to 0.90, suggesting that the items used to measure each variable are highly consistent and reliable.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Reliability and validity.

https://doi.org/10.1371/journal.pone.0330269.t002

Composite Reliability is another measure of internal consistency, similar to Cronbach’s Alpha, but it takes into account the loadings of each item on its corresponding variable. Composite Reliability values above 0.70 are considered acceptable. The Composite Reliability values in this study range from 0.78 to 0.96, indicating high reliability for all variables.

AVE is a measure of convergent validity, which assesses the extent to which the items within each variable are related to the construct they are intended to measure. AVE values above 0.50 are considered satisfactory, as they indicate that the variable explains more than half of the variance in its items. The AVE values in this study range from 0.53 to 0.81, demonstrating good convergent validity for all variables.

The specific variables in the study demonstrate strong reliability and validity. Learning Outcomes (LO) has a Cronbach’s Alpha of 0.77, a Composite Reliability of 0.96, and an AVE of 0.65, suggesting that the items used to measure learning outcomes are reliable and valid. Prior Domain-Specific Knowledge (PDSK) has a Cronbach’s Alpha of 0.72, a Composite Reliability of 0.76, and an AVE of 0.81, indicating that the items used to measure prior domain-specific knowledge are reliable and valid. Prior Experience with Technology-Mediated Learning (PETML) has a Cronbach’s Alpha of 0.83, a Composite Reliability of 0.78, and an AVE of 0.53, suggesting that the items used to measure prior experience with technology-mediated learning are reliable and valid. Finally, Self-Efficacy (SE) has a Cronbach’s Alpha of 0.90, a Composite Reliability of 0.91, and an AVE of 0.74, indicating that the items used to measure self-efficacy are highly reliable and valid.

The cross-loading factors in Table 3 demonstrate a clear pattern of high loadings on the intended constructs and lower loadings on the other constructs. For example, the items measuring Learning Outcomes (LO1 to LO4) have high loadings on the LO construct, ranging from 0.72 to 0.95, and lower loadings on the other constructs. Similarly, the items measuring Prior Domain-Specific Knowledge (PDSK1 to PDSK3) have high loadings on the PDSK construct, ranging from 0.83 to 0.98, and lower loadings on the other constructs. It should be noted that MLTA6 was eliminated from the analysis due to its low loading factor of 0.23 on the MLTA construct and higher cross-loadings on other variables, which indicated poor discriminant validity. This pattern is consistent across all the constructs, including Prior Experience with Technology-Mediated Learning (PETML), Self-Efficacy (SE), and Mobile Learning Technology Adoption (MLTA).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Cross-loading factors.

https://doi.org/10.1371/journal.pone.0330269.t003

The strong loadings of items on their intended constructs and the relatively low loadings on other constructs provide evidence of good discriminant validity in the measurement model. This means that the constructs in the model are distinct from one another and that the items used to measure each construct are effectively capturing the intended concept. Furthermore, the high loadings of items on their intended constructs suggest that the items are converging on the same underlying concept, providing evidence of convergent validity. Convergent validity refers to the extent to which items within a construct are related to one another and collectively measure the same concept.

To further establish discriminant validity, the Fornell-Larcker criterion was applied, which requires that the square root of each construct’s AVE should be greater than its correlations with other constructs. Table 4 presents the results of this assessment, where all diagonal values (square roots of AVE) exceed the corresponding off-diagonal correlation values, confirming discriminant validity. Additionally, the Heterotrait-Monotrait (HTMT) ratio was examined as a more stringent test for discriminant validity. All HTMT values were below the conservative threshold of 0.85, with the highest ratio being 0.72 between Learning Outcomes and Prior Domain-Specific Knowledge, further supporting the discriminant validity of the measurement model. These results, combined with the cross-loading analysis, provide robust evidence that each construct in the model is distinct and captures unique variance not explained by other constructs.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Fornell-Larcker criterion for discriminant validity assessment.

https://doi.org/10.1371/journal.pone.0330269.t004

To assess potential multicollinearity among the predictor constructs, we examined the Variance Inflation Factor (VIF) values provided by SmartPLS. All VIF values were below the conservative threshold of 3.3 recommended for PLS-SEM analysis [50]. Specifically, the VIF values ranged from 1.84 to 2.97 across all predictor constructs (PDSK = 1.84, PETML = 2.43, SE = 2.97), confirming that multicollinearity does not pose a threat to the validity of our structural model results.

Descriptive statistics

Table 5 summarizes the demographic information of participants in this study. In terms of age range, the largest difference was seen in the 22–24 category, where 48% of the Russian participants fell compared to only 15% of the Chinese participants. The gender distribution was fairly similar between the two groups, with both having slightly more female participants. When it came to musical experience, there were notable differences. While 42% of Chinese participants had 15–20 years of experience, only 33% of Russian participants fell in that category. On the other hand, while only 18% of Chinese participants had 5–10 years of experience, 43% of Russian participants fell in that category. The two groups also differed in terms of piano experience, with more Russian participants (46%) having 5–10 years of experience compared to only 8% of Chinese participants.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Demographic information of participants in the study, divided by country.

https://doi.org/10.1371/journal.pone.0330269.t005

Table 6 presents the descriptive statistics for all study variables across the complete dataset and by country subgroups. Initially, 258 participants completed the survey; however, 18 responses were excluded from the analysis due to incomplete surveys with excessive missing data or response patterns indicating lack of engagement (such as providing identical responses to all questions), resulting in a final sample of 240 participants. Missing data for the retained responses were handled using listwise deletion in SmartPLS, ensuring that only complete cases were included in the PLS-SEM analysis to maintain data integrity and model reliability. The complete dataset (N = 240) shows that participants reported moderate to high levels across all constructs, with means ranging from 3.58 to 3.91 on the 5-point Likert scale. Prior Domain-Specific Knowledge had the highest mean (M = 3.91, SD = 0.82), indicating that participants possessed substantial piano-related knowledge and experience. Mobile Learning Technology Adoption showed the lowest mean (M = 3.58, SD = 0.93), suggesting room for improvement in MLT integration among piano learners.

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. Descriptive statistics of study variables.

https://doi.org/10.1371/journal.pone.0330269.t006

Examining the cultural differences, Chinese students demonstrated higher means across all variables compared to their Russian counterparts. Most notably, Chinese students reported higher Prior Experience with Technology-Mediated Learning (M = 3.78, SD = 0.84) compared to Russian students (M = 3.52, SD = 0.93), and higher Mobile Learning Technology Adoption (M = 3.72, SD = 0.87 vs. M = 3.44, SD = 0.98). These differences align with the stronger path coefficients observed in the structural model for Chinese students.

Measurement invariance assessment.

To ensure the validity of cross-cultural comparisons between Chinese and Russian samples, we conducted a Measurement Invariance of Composite Models (MICOM) analysis following the three-step procedure recommended by Henseler et al. (2016).

As shown in Table 7, Step 1 (configural invariance) was established for all constructs, with c-values ranging from 0.998 to 1.000, all exceeding the recommended threshold of 0.70. Step 2 (compositional invariance) was also confirmed for all constructs, as the correlation values (0.994–0.999) exceeded their respective 5% quantiles (0.992–0.999). As presented in Table 7, Step 3 revealed partial invariance for all constructs, with mean and variance differences falling outside the ± 0.30 threshold, indicating that while the constructs measure the same concepts across cultures, there are differences in mean values and variances between the two samples.

Download:

• PNG
  larger image
• TIFF
  original image

Table 7. MICOM analysis results for Chinese vs. Russian samples.

https://doi.org/10.1371/journal.pone.0330269.t007

The establishment of configural and compositional invariance (Steps 1 and 2) confirms that our model can be meaningfully compared across Chinese and Russian samples, supporting the validity of our cross-cultural analysis. The partial invariance in Step 3 suggests cultural differences in construct levels, which is expected and does not compromise the validity of our structural model comparisons.

Structural model evaluation (Hypothesis testing)

Table 8 presents the results of the hypothesis testing for the proposed model across three datasets: the complete dataset, Chinese students, and Russian students. Starting with the complete dataset, the results provide strong support for all the hypothesized relationships. The path coefficient between Prior Domain-Specific Knowledge (PDSK) and Mobile Learning Technology Adoption (MLTA) is 0.547 (95% CI [0.41, 0.69]), with a t-value of 4.612 and a p-value of 0.000. The effect size (f²) of 0.26 indicates a medium effect. This suggests that PDSK has a significant positive effect on MLTA, supporting hypothesis H1.

Download:

• PNG
  larger image
• TIFF
  original image

Table 8. Results of hypotheses testing.

https://doi.org/10.1371/journal.pone.0330269.t008

Similarly, the path coefficient between Prior Experience with Technology-Mediated Learning (PETML) and MLTA is 0.638 (95% CI [0.50, 0.78]), with a t-value of 6.547 and a p-value of 0.000. The effect size of 0.35 indicates a large effect, providing strong support for hypothesis H2. The relationship between Self-Efficacy (SE) and MLTA is also significant, with a path coefficient of 0.639 (95% CI [0.49, 0.78]), a t-value of 4.358, and a p-value of 0.000. The effect size of 0.35 indicates a large effect, confirming hypothesis H3. Finally, the path coefficient between MLTA and Learning Outcomes (LO) is 0.541 (95% CI [0.40, 0.68]), with a t-value of 5.289 and a p-value of 0.000. The effect size of 0.27 indicates a medium effect, supporting hypothesis H4.

Moving on to the Chinese students’ dataset, the results are consistent with the complete dataset, providing support for all the hypotheses. The path coefficients for the relationships between PDSK and MLTA (β = 0.595, 95% CI [0.44, 0.75], f² = 0.31), PETML and MLTA (β = 0.759, 95% CI [0.61, 0.91], f² = 0.50), SE and MLTA (β = 0.660, 95% CI [0.51, 0.81], f² = 0.37), and MLTA and LO (β = 0.588, 95% CI [0.45, 0.73], f² = 0.32) are all significant, confirming hypotheses H1, H2, H3, and H4 in the context of Chinese students. Notably, the effect of PETML on MLTA is particularly strong in this group, with a large effect size of 0.50.

Similarly, the results for the Russian students’ dataset support all the hypothesized relationships. The path coefficients between PDSK and MLTA (β = 0.658, 95% CI [0.49, 0.83], f² = 0.37), PETML and MLTA (β = 0.640, 95% CI [0.49, 0.79], f² = 0.35), SE and MLTA (β = 0.755, 95% CI [0.60, 0.91], f² = 0.49), and MLTA and LO (β = 0.469, 95% CI [0.33, 0.61], f² = 0.21) are all significant, providing evidence for hypotheses H1, H2, H3, and H4 in the context of Russian students. It’s worth noting that the effect of SE on MLTA is particularly strong in this group, with a large effect size of 0.49.

While the overall pattern of results is consistent across the three datasets, there are some differences in the magnitude of the path coefficients and effect sizes. For example, the effect of PETML on MLTA is stronger for Chinese students (β = 0.759, f² = 0.50) compared to the complete dataset (β = 0.638, f² = 0.35) and Russian students (β = 0.640, f² = 0.35). Similarly, the effect of SE on MLTA is stronger for Russian students (β = 0.755, f² = 0.49) compared to the complete dataset (β = 0.639, f² = 0.35) and Chinese students (β = 0.660, f² = 0.37). These differences suggest that the relative importance of the antecedents of MLTA may vary across different cultural contexts.

The confidence intervals provide additional insight into the precision of our estimates. For instance, the narrower confidence interval for the PETML to MLTA relationship in the Chinese sample ([0.61, 0.91]) compared to the Russian sample ([0.49, 0.79]) suggests that we can be more certain about the strength of this relationship in the Chinese context.

Table 9 presents a comparison of the path coefficients between the Chinese and Russian student samples, providing insights into the potential differences in the relationships between the antecedents of Mobile Learning Technology Adoption (MLTA) and its impact on Learning Outcomes (LO) across the two cultural contexts. The first column of the table lists the hypothesized relationships, while the second and third columns display the path coefficients for the Chinese (βC) and Russian (βR) student samples, respectively. The fourth column shows the differences between the path coefficients (βC-βR), and the fifth column presents the p-values associated with these differences.

Download:

• PNG
  larger image
• TIFF
  original image

Table 9. Comparison of path coefficients between Chinese and Russian students.

https://doi.org/10.1371/journal.pone.0330269.t009

Starting with the relationship between Prior Domain-Specific Knowledge (PDSK) and MLTA, the path coefficient for the Chinese student sample (βC = 0.595) is slightly lower than that for the Russian student sample (βR = 0.658), resulting in a difference of −0.063. However, the p-value associated with this difference is 0.428, indicating that the difference is not statistically significant at the conventional 0.05 level. This suggests that the effect of PDSK on MLTA is relatively consistent across the two cultural contexts.

Moving on to the relationship between Prior Experience with Technology-Mediated Learning (PETML) and MLTA, the path coefficient for the Chinese student sample (βC = 0.759) is higher than that for the Russian student sample (βR = 0.640), with a difference of 0.119. The p-value associated with this difference is 0.089, which is lower than the conventional 0.05 level but higher than the more stringent 0.01 level. This suggests that there may be a marginal difference in the effect of PETML on MLTA between the two cultural contexts, with the effect being somewhat stronger for Chinese students.

The relationship between Self-Efficacy (SE) and MLTA shows a similar pattern to that of PDSK and MLTA. The path coefficient for the Chinese student sample (βC = 0.660) is lower than that for the Russian student sample (βR = 0.755), with a difference of −0.095. However, the p-value associated with this difference is 0.429, indicating that the difference is not statistically significant. This suggests that the effect of SE on MLTA is relatively consistent across the two cultural contexts.

Finally, the relationship between MLTA and LO shows a slightly different pattern. The path coefficient for the Chinese student sample (βC = 0.588) is higher than that for the Russian student sample (βR = 0.469), with a difference of 0.119. However, the p-value associated with this difference is 0.327, indicating that the difference is not statistically significant. This suggests that the effect of MLTA on LO is relatively consistent across the two cultural contexts.

These results presented in Table 9 provide a nuanced understanding of the similarities and differences in the relationships between the antecedents of MLTA and its impact on LO across Chinese and Russian student samples. While the overall pattern of results is relatively consistent, with no statistically significant differences found, there is a marginal difference in the effect of PETML on MLTA between the two cultural contexts. These findings highlight the importance of considering cultural factors when examining the adoption and effectiveness of mobile learning technologies in music education and suggest that the proposed model may have broad applicability across different cultural contexts, with some potential variations in the relative importance of specific antecedents.

Model fit and predictive relevance.

To assess the overall fit and predictive power of our model, we examined several key indices across all three datasets (Complete, Chinese, and Russian). The coefficient of determination (R²) values for the endogenous constructs provides an indication of the model’s predictive accuracy. For the complete dataset, the R² value for Mobile Learning Technology Adoption (MLTA) was 0.56, indicating that 56% of the variance in MLTA is explained by its predictors (PDSK, PETML, and SE). The R² value for Learning Outcomes (LO) was 0.29, suggesting that MLTA explains 29% of the variance in LO. These R² values indicate a moderate to substantial level of predictive accuracy according to Hair et al. (2023). For the Chinese sample, the R² values were slightly higher, with MLTA at 0.60 and LO at 0.35. The Russian sample showed similar results, with R² values of 0.57 for MLTA and 0.22 for LO.

To assess the model’s predictive relevance, we calculated the Stone-Geisser Q² value using the blindfolding procedure. Q² values larger than zero indicate the path model’s predictive relevance for a particular construct. In the complete dataset, the Q² values were 0.36 for MLTA and 0.19 for LO, both exceeding zero and thus supporting the model’s predictive relevance. Similar results were found in the Chinese (Q²MLTA = 0.39, Q²LO = 0.23) and Russian (Q²MLTA = 0.37, Q²LO = 0.15) samples.

Finally, we assessed the model’s overall fit using the Standardized Root Mean Square Residual (SRMR). The SRMR is an absolute measure of fit, with values less than 0.08 considered to represent good fit. Our model demonstrated good fit across all datasets, with SRMR values of 0.054 for the complete dataset, 0.050 for the Chinese sample, and 0.057 for the Russian sample. These results collectively suggest that our model has good predictive power and fits the data well across all three datasets, providing further support for the validity of our findings.

As shown in Table 10, all relationships demonstrated meaningful effect sizes exceeding the 0.15 threshold for medium effects. Prior Experience with Technology-Mediated Learning and Self-Efficacy consistently showed large effects (f² ≥ 0.35) across all samples, with particularly strong effects among Chinese students for PETML (f² = 0.50) and Russian students for SE (f² = 0.49). Prior Domain-Specific Knowledge exhibited medium to large effects (f² = 0.26–0.37), while MLTA’s impact on Learning Outcomes showed consistent medium effects (f² = 0.21–0.32) across all cultural contexts.

Download:

• PNG
  larger image
• TIFF
  original image

Table 10. Stone-Geisser Q² values and Cohen’s f² effect sizes for MLT adoption model.

https://doi.org/10.1371/journal.pone.0330269.t010

The Stone-Geisser Q² values presented in Table 10 demonstrate the model’s predictive relevance across all datasets. All Q² values exceeded zero, confirming predictive relevance for both endogenous constructs. MLTA showed strong predictive relevance with Q² values ranging from 0.36 to 0.39, while Learning Outcomes exhibited moderate predictive relevance with Q² values between 0.15 and 0.23. Chinese students demonstrated the highest predictive relevance for both constructs (Q²MLTA = 0.39, Q²LO = 0.23).

Theoretical contributions

This study makes several important theoretical contributions to the understanding of adoption of mobile learning technology and its impact on learning outcomes in the context of piano education. By proposing and empirically validating a comprehensive model that integrates key antecedents of mobile learning technology adoption, such as prior domain-specific knowledge, prior experience with technology-mediated learning, and self-efficacy, this study extends the existing literature on technology acceptance and music education.

First, the study contributes to the technology acceptance literature by demonstrating the applicability of established constructs, such as prior experience and self-efficacy, to the specific context of mobile learning technology adoption in piano education. While previous research has examined these constructs in various educational settings [44,45], this study specifically focuses on their role in the adoption of mobile learning technologies for piano learning. By doing so, it provides a more nuanced understanding of how these factors influence learners’ attitudes and behaviors towards mobile learning technologies in the domain of music education.

Second, the study extends the literature on music education by introducing the concept of prior domain-specific knowledge as a key antecedent of mobile learning technology adoption in piano learning. While prior research has acknowledged the importance of prior musical knowledge and skills in music learning [71,72], this study specifically examines its role in the adoption of mobile learning technologies. By demonstrating the significant positive effect of prior domain-specific knowledge on mobile learning technology adoption, this study highlights the need for music educators and researchers to consider learners’ pre-existing knowledge and skills when designing and implementing mobile learning interventions in piano education.

Third, the study contributes to the growing body of literature on mobile learning in music education by providing empirical evidence for the positive impact of mobile learning technology adoption on piano learning outcomes. While previous studies have explored the potential benefits of mobile learning technologies in various areas of music education, this study specifically focuses on piano learning and provides a comprehensive model that links mobile learning technology adoption to improved learning outcomes. By doing so, it offers a theoretical foundation for future research on the effectiveness of mobile learning interventions in piano education and highlights the need for music educators to leverage mobile technologies to support learners’ skill development and knowledge acquisition.

Finally, the study contributes to the understanding of cross-cultural differences in mobile learning technology adoption and its impact on learning outcomes in piano education. By comparing the results across Chinese and Russian student samples, the study demonstrates the generalizability of the proposed model and highlights the consistency of the relationships between the antecedents of mobile learning technology adoption and its impact on piano learning outcomes across different cultural contexts. This finding suggests that the key factors influencing mobile learning technology adoption and effectiveness in piano education may be largely universal, despite potential cultural differences. However, the study also identifies a marginal difference in the effect of prior experience with technology-mediated learning on mobile learning technology adoption between the two cultural contexts, underlining the need for further research to explore the role of cultural factors in shaping the relative importance of specific antecedents.

Practical implications

Building on our findings, several practical recommendations emerge for educators and MLT designers in piano education. Our study reveals that students’ existing piano knowledge significantly influences their MLT adoption, suggesting educators should assess musical backgrounds before implementing these technologies. Students with stronger musical foundations readily embrace MLTs and can access advanced features immediately, while those with limited prior knowledge may require foundational musical instruction before technological integration.

Prior experience with educational technology emerged as a critical adoption factor, particularly among Chinese students. Educators should identify students with limited technology-mediated learning backgrounds and provide additional support during initial MLT implementation. Conversely, students comfortable with educational platforms can quickly advance to sophisticated applications.

Self-efficacy in MLT use significantly affects adoption, especially among Russian students. Building students’ confidence through structured early success experiences becomes essential, starting with simpler applications before progressing to complex tools. Early positive experiences prevent discouragement and promote continued engagement.

The demonstrated positive impact of MLT adoption on learning outcomes justifies institutional investment in thoughtful implementation. However, success requires addressing individual readiness factors rather than uniform technology provision. MLT developers should design applications that accommodate varying musical knowledge levels and build user confidence through adaptive interfaces and encouraging feedback mechanisms. Cultural differences suggest that effective implementation strategies may vary across educational contexts, requiring tailored approaches rather than universal solutions.

Limitations and future research recommendations

While this study makes significant contributions to the understanding of mobile learning technology adoption and its impact on piano learning outcomes, there are several limitations that should be acknowledged and addressed in future research.

On the one hand, the study examined a limited set of antecedents of mobile learning technology adoption, focusing on prior domain-specific knowledge, prior experience with technology-mediated learning, and self-efficacy. Future research could explore additional factors that may influence the adoption and effectiveness of mobile learning technologies in piano education, such as learning style preferences, motivation, or social support. Incorporating a broader range of antecedents would provide a more comprehensive understanding of the complex interplay of factors that shape learners’ attitudes and behaviors towards mobile learning technologies in piano education.

On the other hand, the study did not examine the specific features or design elements of mobile learning technologies that may contribute to their effectiveness in piano education. Future research could investigate the impact of different instructional strategies, feedback mechanisms, or gamification elements on learners’ engagement, motivation, and performance. This would provide valuable insights for mobile learning technology developers and music educators, enabling them to optimize the design of mobile learning solutions for piano education.

Additionally, the study relies on self-reported data from students, which may introduce biases such as social desirability or inaccurate self-assessment. Future research could benefit from combining self-reported measures with objective assessments of learning outcomes, such as tracking actual performance improvements over time using MLTs, to provide a more accurate and comprehensive evaluation of MLT effectiveness.

Furthermore, the study was limited to participants aged 18–26 years, representing a relatively narrow age demographic that grew up with digital technologies. When it comes to the use of Mobile Learning Technologies, a wider age range would likely significantly change the results, as different age groups may have varying levels of comfort, familiarity, and adoption patterns with mobile technologies. Future research should examine MLT adoption across broader age ranges to understand how generational differences in technology exposure and digital literacy influence the relationships identified in this study.

Finally, while this sampling method provides a good representation of students within the selected institutions, it is important to acknowledge the limitations in generalizing the findings to all music students in China and Russia. The selected universities, although prestigious, may have unique characteristics in terms of curriculum, student demographics, and technology adoption that might not be typical of all music education programs in these countries. Future research could benefit from including a wider range of institutions across different regions to provide a more comprehensive representation of music students in China and Russia.