Introduction

English language education in developing countries such as Malaysia has frequently been highlighted as an important qualification to transform the country into a high-income nation (Renganathan, 2023). Hence, English is a compulsory subject taught in primary and secondary schools for 11 years as a second language (Nik-Fauzi et al., 2022) and used as a medium of instruction at the tertiary level in most higher learning institutions. Likewise, due to the exponential growth of technology, Elaish et al. (2023) describe that using technologically assisted language-learning tools has become essential to teaching English as a second language (ESL) as learning behaviour has shifted from passive to active in recent years. Furthermore, technology adoption to support language learning is imperative to cater to the millennials’ learning needs (Shadiev & Yang, 2020). Even though this has simultaneously created new opportunities to blend digital technologies for language learning (Buragohain et al., 2023), the pandemic entrusted educators and students to digitalise learning at a whole new level by adopting a distance learning approach. Hence, online learning was no longer merely an option but a necessity to enhance learning with possible merits (Chung et al., 2020) as it gained prominence in education, envisaging higher use intention post-pandemic (Al-Hamad et al., 2021). In Malaysia, pre-pandemic online learning has been regarded as transformative for ESL by integrating tools such as YouTube, Google Classroom, and WhatsApp to complement face-to-face teaching (Haleman & Yamat, 2021; Rahman, 2020). However, post-pandemic, ESL learners have now experienced a higher degree of using online tools for videoconferencing and collaboration, such as Zoom, Skype, and WebEx, as well as tools to manage and communicate online to facilitate learning (Nik-Fauzi et al., 2022). According to Chew and Ng (2021) and Krishan et al. (2020), ESL learners benefited from synchronously using online resources such as Google Translate and dictionaries during virtual classes where the instant access to these tools improved their language skills and confidence in communicating in English. Simultaneously, it also aided constructive and self-directed learning abilities in improving basic language skills, especially for reading, writing, and vocabulary acquisition through tools such as Typely and Grammarly (Buragohain et al., 2023). Consequently, the purpose of these tools shifted from merely an option to complement classroom learning dictated by teaching needs to a critical access point for learning and communicating independently.

In retrospect, modicum studies have addressed ESL students’ acceptance and efficacy in independently using these technologies (Zaidan et al., 2021). Likewise, the unexpected switch to online education during the pandemic also prompted similar concerns about how it relates to digital literacy (Alfadda & Mahdi, 2021) and the psychological challenges involved in language learning (Karuppannan & Mohammed, 2020). Uztosun (2020) highlighted the importance of considering these characteristics as a considerable amount of language learning skills, while discrete, highly influences ESL learners’ performances. Furthermore, in Malaysia, research on ESL online learning strategies often focuses on the technical angle while neglecting students’ opinions, which is vital in improving technology assimilation (Nik-Fauzi et al., 2022). Undoubtedly, students’ perception, attitude, and needs towards online learning has also shifted post-pandemic (Alfadda & Mahdi, 2021) which warrants exploration to provide critical insights to maximise learning outcomes (Zapata-Cuervo et al., 2022), especially for teaching and learning ESL (Buragohain et al., 2023). Therefore, Mohtar and Yunus (2022) asserted that the transition makes examining ESL students’ perceptions, acceptance, and intention of online learning necessary.

Therefore, Haleman and Yamat (2021) suggested expanding the technology acceptance model (TAM) (Davis, 1985) by considering language-learning factors as external factors. TAM is frequently utilised in investigating users’ technology acceptance by shedding light on how technology interaction influences attitude (ATT) and behavioural intention (BI), mainly based on the perceived ease of use (PEOU) and perceived usefulness (PU) of the said technology. TAM theoretically focuses on the effects of technology interaction but could be adapted to predict learning behaviour, intention, and attitude due to technology use (Kumar et al., 2020). Next, regarding external factors, we first considered experience with online learning in higher education post pandemic, as Lazar et al. (2020) and Wei (2022) suggested. As online learning experience rendered familiarity with education technology and online learning anxieties, these two constructs were deemed necessary to predict future intention. According to Hanafiah and Aziz (2022), in Malaysia, very few studies emphasise the psychological attributes of online language learning caused by the pandemic. Hence, we firstly considered online learning anxiety, as suggested by Wang and Zhang (2021), and secondly, ESL learning confidence, as suggested by Côté and Gaffney (2021). ESL learners’ confidence in learning is a primary concern for Malaysian undergraduates (Mohamad, 2020) and exploring this perspective is vital in understanding ESL learners’ motivation to adapt to online learning postpandemic (Siok et al., 2023). Therefore, this study explores the acceptance and intention to use online learning to learn English by answering the following research questions:

• RQ1: Which factors significantly influence attitude and behavioural intention to use online learning to learn English based on TAM?

• RQ2: How do external factors, namely confidence in English, online learning anxiety, and familiarity with education technology, influence attitude and behavioural intention to use online learning to learn English?

The following section discusses how the conceptual model was hypothesised by incorporating these factors.

Hypotheses development

The hypotheses developed in this study are described in terms of the endogenous and exogenous variables used to predict utilisation attitude and behavioural intention per the general extended technology acceptance model for e-learning (GETAMEL) (Abdullah & Ward, 2016). GETAMEL considers the TAM variables as internal constructs of the model and external constructs as antecedents of the primary internal constructs (PU and PEOU) (Jiang et al., 2021).

External constructs

Confidence in English (CONF)

Confidence is one of the main factors influencing language acquisition (Krashen, 1981). ESL learners tend to be passive when learning online, which can be attributed to low confidence in language proficiency (Putri, 2023) when communicating with their instructor or when singled out in a classroom environment to provide feedback. Nevertheless, integrating technology in the language classroom during the pandemic has been deemed transformative in learning experiences and levels of confidence (Burk, 2021). Therefore, we theorised CONF as an external variable that may influence usefulness and ease of use in the context of online language learning, as also suggested by Haleman and Yamat (2021):

• H6: CONF will significantly influence PU.

• H7: CONF will significantly influence PEOU.

Online Learning Anxiety (ANX)

One of the most critical emotional aspects influencing university students’ English language acquisition due to the pandemic is online learning anxiety: environmental anxiety attributed to information overload, technical factors, and learning difficulties (Wang & Zhang, 2021). According to Zapata-Cuervo et al. (2022), online learning during the pandemic has increased anxiety due to learning challenges, which Abdelwahed et al. (2022) reflect as difficulties in access, technical support, and a lack of technical skills required for learning a language. Hence, anxiety is influenced by ease of use as also suggested by Abdullah and Ward (2016). Additionally, as anxiety is frequently associated with usefulness, the experience of utilising technologies may significantly impact adoption (Binyamin, 2019). Tapsuri and Polyiem (2022) explained that ANX could also be attributed to individual characteristics reflecting learning tool preferences due to their recognised usefulness. Therefore, we hypothesised the following relationships:

• H8: ANX will significantly influence PU.

• H9: ANX will significantly influence PEOU.

Familiarity with Education Technology (EdTech)

Students’ technological competency in learning is determined by the extent and quality of their digital learning experience, which indirectly influences their online learning attitude (Lazar et al., 2020). Buragohain et al. (2023) stated that limited learning technology exposure might influence language pedagogical practises and online learning attitudes, which Alfadda and Mahdi (2021) claim are associated with TAM factors. Likewise, although they had exposure to online learning due to the pandemic, Malaysian ESL undergraduates are not extensively exposed to these technologies to signify strong familiarity with education technologies (Hasnan & Mohin, 2021). Jiang et al. (2021) claimed that familiarity did not influence PU or PEOU in language learning; however, we hypothesise, based on differing views, the following relationships:

• H10: EdTech will significantly influence PU.

• H11: EdTech will significantly influence PEOU.

• H12: EdTech will significantly influence ATT.

Accordingly, based on the hypothesised relationships, we proposed the conceptual model as represented in Figure 1.

Materials and methods

Analysis and findings

Measurement model

Measurement model analysis determines the model’s reliability and validity to observe the relationship between the data and variables. The findings (Table 1) revealed that all reliability values represented by the indicator reliability, composite reliability, and internal consistency were below the threshold value suggested by Hair et al. (2019), signifying that the data had internal consistency.

Next, the convergent validity determined, based on the average variance extracted (AVE), signifies that each construct can account for the variation in its elements. Likewise, all Heterotrait-Monotrait (HTMT) values were below the threshold value of 1.00 (Henseler et al., 2015) (Table 2), revealing that respondents could easily differentiate between the constructs. Hence, the measurement model was also acceptable.

Structural model

The relationships between the variables and the hypothesised relationships were measured using a structural model. First, the model fit was determined by the standard root mean square residual (SRMR) value, indicating a value of 0.057 which was below the cut-off value of 0.08 (Henseler et al., 2016). Next, the variance inflation factor (VIF) analysis (Table 3) which determines items’ correlation with each other, indicated that all values were below the threshold value of 5.0 (Henseler et al., 2016). As both these values were established, the analysis of path coefficients (β) used to determine the correlation and strengths between variables was performed (Wong, 2019), followed by bootstrapping resampling of 5,000 that indicated significant relationships when t-values were above 1.96 (Hair et al., 2019). Hence, the results of all hypothesised relationships based on β, t-value, confidence intervals, and effect size (f2) are presented in Table 3, where non-significant relationships are highlighted in red, as shown in Figure 2.

The analysis showed that all hypotheses failed to be rejected except for H₁₀ (β=0.067, t=1.834, =0.067, f²=0.009) and H₁₂ (β=0.048, t=1.228, =0.220, f²=0.005) (Table 3).

Therefore, familiarity with education technology post-pandemic for learning English did not influence attitude and perceived usefulness, but the ease of use of such platforms (H₁₁: β=0.308, t=7.374, =0.000, f²=0.194). According to Cohen (2013), f² values between 0.020 - 0.150 are weak, 0.150 to 0.350 are medium, and values equal to or larger than 0.350 are large effect size. H₅ (β=0.761, t=24.191, =0.000, f²=1.148) had the highest effect size, followed by H₃ (β=0.577, t=13.972, =0.000, f²=0.409) where both were considered as having large effect size. Likewise, H₂ (β=0.366, t=6.698, =0.000, f²=0.152), H4 (β=0.491, t=7.366, p=0.000, f²=0.246) and H₉ (β=0.397, t=7.083, =0.000, f²=0.242) had medium effect size.

Next, the coefficient determination (R²) indicated that the model could explain 81.9% of BI, 72.6% of ATT, 72.2% of PU, and 65.9% of PEOU. Hair et al. (2019) described R² values of 0.75, categorised as strong, 0.50 as moderate, and 0.25 as weak. Based on the findings, it could be determined that the in-sample predictive power is considered strong for BI (0.819) and ATT (0.726). According to Benitez et al. (2020), a high value can be expected when the phenomenon is already well understood.

However, to determine the predecessors with a strong overall effect (i.e., those with reasonably high importance for the target construct) and importance, the importance-performance map analysis (IPMA) was conducted (Ringle & Sarstedt, 2016).

Based on the findings as reflected in Table 4 and Figure 3, the importance (total effects) for BI is mainly determined by ATT (0.773) and PEOU (0.691) and less so by CONF (0.253) and EdTech (0.264). IPMA helps to identify areas of improvement in predicting BI, and the performance value of EdTech (72.318), ATT (68.323), and PEOU (67.966) and the area of improvement in future studies.

Discussion

The findings indicated that the hypothesised model could predict 81.9% of behavioural intention and 72.6% of attitude, which reflects strong in-sample predictive power. Referring to RQ1 for internal constructs, attitude was the strongest predictor of intention in using online learning with a strong impact. These results align with those obtained by Mailizar et al. (2021). Likewise, ease of use was pivotal in determining usefulness and attitude with a high effect corresponding to Annamalai et al. (2021) findings, yet the association between attitude and usefulness differed as a significant but weak effect was observed. Therefore, this reflects a movement away from the relevance of perceived usefulness towards a more significant influence of ease of use. Although language learners’ intention to use online learning could be heavily influenced by their ability to access information, internet connection, and technical facilities (Baxter, 2020); Rubaai and Khatib (2020) added that the association between simplicity and usefulness could only be confirmed if they perceive easy access to and means of interaction with learning resources, online collaboration, and authentic instructor feedback. Hence, simplicity in accessing and using online tools is vital in determining utilisation, and we theorised that this could be a behaviour inherited due to the convenience of using online resources for vocabulary, phrases, sentences, spelling, and grammar for virtual classes during the pandemic.

Next, to answer the second research question (RQ2), focusing on external constructs, the findings indicated that all external factors were positively associated with ease of use but only reflected a medium effect. Confidence in English and online learning anxiety were found to have a weak relationship with usefulness, while familiarity with education technology did not influence usefulness, as observed by Jiang et al. (2021). Additionally, familiarity also did not influence attitude, which contradicted the findings of Abdullah and Ward (2016). Interestingly, even for external constructs, PEOU still precedes PU. According to Kumar et al. (2020), usefulness is more influential than ease in initial platform adoption and not for continuous use. Moreover, due to the pandemic, the online learning experience may have transformed language learners’ attitude that focuses on ease of use as online learning has become a substantial norm. Furthermore, the pandemic orchestrated a pivotal movement in using learning technologies by acting as a catalyst for digital transformation in higher education (Cazan & Maican, 2023), where online language learning, which previously focused on distributing informative content and conducting online activities that supplement face-to-face instruction (Azlan et al., 2020), is no longer a norm. Conversely, while Malaysian students were marginally prepared for online learning due to a lack of individual learning experiences and technical difficulties pre-pandemic (Chung et al., 2020), mandatory online learning has positively benefited them (Bervell et al., 2022) during the pandemic, where they are motivated to participate and accept the possibilities of language learning online (Mohtar & Yunus, 2022).

One interesting aspect that emerged from the analysis is the stronger association between confidence in English and online learning anxiety towards ease of use compared to usefulness. Confidence in English mirrors the level of learning participation (Ramsa & Mohd-Rawian, 2019), which reduces learning anxiety (Abdous, 2019). Abdullah and Ward (2016) and Burk (2021) explained that higher levels of language learning confidence could be due to exposure to learning tools. While we did not explore this hypothesised relationship, we rationalise this association based on the precedence of ease of use after the initial introduction of a tool for achieving learning goals which we relate to usefulness. Hence, unsurprisingly, the continuous use of online learning tools has rendered familiarity inconsequential towards usefulness, as also highlighted by Binyamin (2019). Our findings also draw similar conclusions to Abdullah and Ward (2016) and Wang and Zhang (2021), suggesting marginal online learning anxiety associations with factors and challenges related to ease of accessing and using online tools in English language acquisition, contradicting the findings by Tapsuri and Polyiem (2022) emphasising usefulness. Nevertheless, our findings also support Putri’s (2023) claims that ESL confidence in learning English online is still a passive experience, and based on the IPMA results, improvement is necessary with online learning experience and anxiety if the goal is to improve intention. Moreover, the results also highlighted the importance of attitude and ease of use towards intention which supports Buragohain et al. (2023) notion that language learning online will continue to acclimate and advance collectively.

Conclusions

Therefore, we conclude that the main factor influencing attitude and intention to use online learning for ESL learners is perceived ease of use. Ease of use was vital not only in arbitrating perceived usefulness but also confidence in English, online learning anxiety, and familiarity with education technology towards the endogenous variables. Likewise, external variables had a stronger association with ease of use than usefulness, where familiarity or experience during the pandemic has rendered inconsequential influence on online learning attitude and usefulness. The findings are vital in transforming ESL language learning online, as understanding motivational factors could maximise future intention (Siok et al., 2023) by aiding educators and stakeholders in identifying ideal instructional interventions to support online language learning. Likewise, the findings highlighted the need to explore how technical factors such as complexity, availability, and accessibility of language tools that reflect ease may influence attitude and intention to use online learning post pandemic.

Some limitations should be considered. First, the findings did not represent different levels of education or specific language learning skills, which warrants further exploration. We also highlight the need to explore factors such as facilitating conditions, behavioural expectations, usability, task–technology fit, habit, and hedonic motivation as the pandemic experience has altered learning experiences and how these variables are associated with ease of use and intention. Next, future studies on the emotional aspect of learning are also recommended, especially considering playfulness and enjoyment in online learning, as we theorised that they may have possible associations with usefulness. Lastly, although the results of this study underline the importance of ease of use, we believe future studies could also consider a mixed-method approach to rationalise the associations between these variables and consider untested language learning variables as mediating or moderating variables. (1)