Exploring how executive function and motivation shape deep and surface learning Free

Higher education (HE) institutions consistently face challenges in preparing students for the workforce and developing their expertise. In the digital era, adopting executive function (EF) skills, such as self-regulation, strategic thinking and motivation, is essential for success, as students engage in complex cognitive and behavioural processes. In online distance education (DE), higher EF capability is linked to a greater use of learning techniques and self-regulation that highlight the increased demands placed on these mental functions in digital and AI-assisted educational settings (Fajardo-Ramos, Chiappe, & Mella-Norambuena, 2025). Students might adopt surface learning strategies to deal with inadequate course material and lack of supporting strategies for EF development and self-regulated learning (Meijs, Gijselaers, Xu, Kirschner, & De Groot, 2021). While improving these skills is critical for enhancing academic performance, progress is often hindered by factors like ability and effort, both linked to students' cognitive skills, motivational capacities and their approach to learning in their studies (Hidi & Harackiewicz, 2000; Zimmerman, 2002). This is especially pertinent in South Africa, where HE institutions are still struggling to deal with a diverse range of student abilities and environmental constraints according to the annual report of the Council on Higher Education (CHE; 2024/2025).

Despite prior research examining influences of motivation and EF on students' learning, little is known about the association on the adoption of a deep or surface learning approach, especially in distance teacher education. Many studies have looked at different facets of HE students in a broad sense. However, research on the association between various forms of motivation and students' EF skills remains limited. Furthermore, there are very few studies on the relation between learning approaches and EFs in South Africa. Therefore, the main research question of this study is: How do EF and motivation shape deep and surface learning among South African student teachers in a DL context? This research is relevant since surface learning approaches are frequently encouraged by the growth of DL- and AI-assisted assessments (Pergantis, Bamicha, Skianis, & Drigas, 2025). This study contributes to existing research by exploring deep and surface learning within a digital learning setting. Moreover, this study incorporates various assessment scales: Executive Skills Questionnaire (ESQ-R), Academic Motivation Scale (AMS-C) and the Study Process Questionnaire (R-SPQ-2F) to understand the learning process in an online, digital context. The results of this study could benefit HE institutions by guiding the development and inclusion of EF and motivation into curricula, which will eventually improve institutional efficiency, success and performance.

Controlling one's thoughts and behaviours to achieve a goal is referred to as executive function (EF), a broad and complex construct (Bunge, 2024). EF serves as the primary control system of the brain, organising and managing the activities vital for performance and learning (Manuhuwa, Snel-de-Boer, de-Graaf, & Fleer, 2024). In addition, EF operations occur in the pre-frontal cortex, a vulnerable part of the human brain and these networks communicate with other brain regions to produce new connections, improving EF (Spreij, Van Tuijl, & Leseman, 2023).

EFs comprise three main competencies, namely cognitive flexibility (CF), a component of thinking that enables a person to solve everyday challenges, adjust to changes in the environment and come up with innovative ideas that spur creativity (Algharaibeh, 2020). When an approach to problem-solving fails, CF is needed to think of or to come up with alternative approaches (Diamond & Ling, 2016) to deal with the particular situation. The second competency refers to working memory (WM), the capacity to execute complex cognitive operations such as reasoning, solving problems and comprehending language (Amzil, 2022). These cognitive processes are evident in activities like listening to lectures, connecting current information with previously learned concepts or establishing causal links between outcomes and events (Diamond & Ling, 2016). Finaly, inhibitory control (IC) enables one to control attention, thoughts and emotions (Bai, Liu, & Su, 2023; Ramos-Galarza, Acosta-Rodas, Bolaños-Pasquel, & Lepe-Martínez, 2019). The presence of IC empowers individuals to modify and determine their responses and actions instead of adhering to traditional habits (Diamond, 2013).

Higher-order EFs develop on the mentioned core functions that includes a) planning: creating, assessing and choosing actions to reach a goal (Bai et al., 2023), b) organisation: being accountable for accessing and arranging essential task components, c) task and self-monitoring: adapting to different approaches and keep track of studies, d) initiation: work independently, e) self-motivation: stay motivated during challenges, f) emotional regulation: regulate reactions (Ramos-Galarza et al., 2019), g) decision-making: enables reasoning and emotions, and h) problem-solving: advanced mental processes such as memory, perception and attention (Pinochet-Quiroz et al., 2022). All related to self-regulated learning, academic achievement and motivation.

Efficient teachers need strong EFs to monitor, assess and reflect on their teaching and make changes as needed. This involves organising, observing, planning and managing their behaviour to accomplish certain goals (Correia & Navarrete, 2017). The good news is, with effort and repetition, EFs can be learned and enhanced (Pergantis et al., 2025).

How students approach learning in HE has been researched for decades. Initially, learning was defined as the total number of test-correct responses. However, in 1976 HE students' reading and processing of academic texts were examined by Marton and Säljö. The authors found two distinct methods for accomplishing this: surface- and deep processing. The two methods more suited to students' learning and studying were modified to deep and surface approaches to learning (Entwistle, McCune, & Scheja, 2006).

Deep learning approaches are associated with high-quality learning outcomes. These approaches emphasise students' active participation in their education and promote a deeper understanding of the content (Masuku, Jili, & Sabela, 2020). Adopting a deep approach to learning indicates an effort to comprehend concepts on a person's own. Every student has a different approach to learning, consisting of motivation and strategy used to complete an activity or find solutions to problems (Biggs, 1991). The main idea of deep motive (intention of learning) is to understand the content (Aderibigbe, 2021), to seek for purpose (Biggs, 1991) and automatically trying to acquire specific information and ensure one gets the overall picture (Biggs & Tang, 2007). Within the deep strategy (process of learning), the aim is to address interest, maximise understanding, actively engage with the content and concentrate on details of the content (Biggs, 1991).

In contrast, a surface approach to learning aims to replicate the subject matter through memorisation (Dolmans, Loyens, Marcq, & Gijbels, 2016). Within this approach, students aim to complete higher-order thinking activities as quickly as possible resulting in substandard work. Students using this strategy rely on memorisation as the best approach for achieving their goal and must balance studying too hard and preventing failure (Biggs, 1991). However, a memorising fact do not only lend itself to a surface approach. When the student memorised the content to provide the impression of understanding, it turns into a surface approach (Biggs & Tang, 2007). In the surface motive approach, the goal is to minimise work and only replicate content to reach goals (Fu & Liu, 2024a). On the other hand, the surface strategy approach uses basic cognitive thinking to complete tasks not worrying about meaning (Biggs & Tang, 2007).

In education, a deep learning approach is preferred to enhance twenty-first century skills in HE students. It is crucial to help students grow their knowledge and skills through a deep learning approach to prepare them for future employment, lifelong learning and social issues (Mthethwa-Kunene, Rugube, & Maphosa, 2021) also reflected in the UN's Sustainable Development Goal 4 (SDG 4). Furthermore, employing a deep approach allows students to grasp the learning content easier, solve problems, think critically, establish how to learn, contextualise learning, use digital tools, gain skills and knowledge and become lifelong learners (Mthethwa-Kunene et al., 2021), all linked to the required twenty-first century skills. Although several factors play a role in student's approach to learning, this study focus mainly on the role of motivation and EF skills in a DL context (Diamond & Ling, 2016; Ryan & Deci, 2020).

People's ability to control their behaviour is heavily impacted by the multifaceted and complicated concept of motivation (Kirkpatrick, Kirkpatrick, & Derakhshan, 2024). Self-determination theory (SDT) is one of the most influential frameworks for understanding human motivation and thriving (Ryan & Deci, 2017). Importantly, and unlike alternative theories that conceptualise motivation as a unitary construct, SDT differentiates various types of motivation along a continuum ranging from amotivation (signifying the lack of intentionality) to extrinsic motivation (associated with a feeling of being coerced or compelled to act) and ultimately to intrinsic motivation as the most autonomous form of motivation. When an individual is intrinsically motivated, behaviours are performed purely out of interest and for the sake of the associated feelings of spontaneous pleasure and enjoyment that results from those behaviours (Ryan & Deci, 2017).

Intrinsic motivation has been described by Deci (1992) as having both dispositional and experiential elements: dispositional elements refer to the intention to keep doing things and experiential elements refer to focused activity involvement, participation and feelings of pleasure, interest and enthusiasm. According to Fairchild, Horst, Finney, and Barron (2005) and Utvær and Haugan (2016), intrinsic motivation comprises three categories: intrinsic motivation to know (find learning something new enjoyable and fulfilling), intrinsic motivation to accomplish (outperforming oneself in studies provides joy) and intrinsic motivation to experience stimulation (finding enjoyment in studying fascinating people). In contrast, extrinsic motivation involves behaviour enacted for reasons that do not stem from inherent pleasure and satisfaction (Ryan & Deci, 2020). Fairchild et al. (2005) identify three sub-components of extrinsic motivation: external regulation (controlled behaviour due to outside rewards and punishments), introjected regulation (preventing guilty feelings or pursuing fulfilment in life) and identified regulation (a person intentionally valuing the purpose of the task).

Finally, amotivation refers to a lack of motivation, which is neither intrinsic nor extrinsic (Can, 2015; Fairchild et al., 2005; Vallerand et al., 1992). In this situation, people either believe that they cannot effectively perform the actions that are needed to attain outcomes, or it may also stem from a lack of interest simply because they attach no value to it. In some cases, amotivation may also be a case of resistance to influence, which can result either in nonaction or in oppositional behaviour to defy demands (Ryan & Deci, 2017). Those who lack motivation believe that circumstances beyond their control are to blame for their actions (Vallerand et al., 1992). Essentially, in DE, students who feel that they do not comprise the necessary skills, do not have a frame of reference to understand the relevance of the study material, or simply have behavioural challenges may be prone to amotivation. It is therefore vital that online learning environments are intentionally designed to emphasise the relevance and real-world applicability of the study material.

According to a SDT framework (Ryan & Deci, 2017), motivation stems from the satisfaction of three fundamental psychological needs – relatedness, competence and autonomy – as essential drivers towards more autonomously motivated behaviour. Moreover, the degree to whether a student's fundamental psychological needs are met or not is determined by both psychological and contextual factors. Vansteenkiste, Ryan and Soenens (2020) describe the three psychological needs in which the feeling of satisfaction, desire and eagerness is referred to as autonomy. When students are satisfied, they will be motivated and feel that they are in control of their studies and learning choices. Relatedness is fulfilled when one connects with and feels important to others, signifying the feeling of comfort, connection and compassion. The sensation of efficacy and mastery is what competence is all about. When one successfully participates in activities and has the chance to apply and expand their knowledge and abilities, it is satisfied.

In addition to the psychological needs, self-efficacy, resilience and attitudes impact students' confidence in their capacity to cope and thrive. Students who have less self-efficacy regulation would be more stressed and, as a result, less motivated (Abdolrezapour, Jahanbakhsh Ganjeh, & Ghanbari, 2023). Because students are unfamiliar with the online setting, self-efficacy is a key influence in online learning (Shen, Cho, Tsai, & Marra, 2013; Zimmerman & Kulikowich, 2016). Resilience addresses students' capacity to overcome challenges and achieve success in their studies (Vance, Pendergast, & Garvis, 2015). According to SDT, the satisfaction of the three basic psychological needs, i.e. relatedness, autonomy and competence, improves well-being and strengthens inner resources related to resilience. Motivation and learning are also significantly impacted by students' attitudes (Kirkpatrick et al., 2024), interests (Hidi & Harackiewicz, 2000) and accomplishment goals (Fairchild et al., 2005). Though the motive behind these goals is crucial for success, positive attitudes increase participation, ongoing interest generates intrinsic motivation and students' goals encourage deeper learning.

One important contextual aspect involves lecturer support, which has been reported as a significant impact of student motivation. According to Ryan and Deci, 2020, lecturers supporting student autonomy provide them with valuable options and opportunities to control their studies and consider their viewpoints, and they encourage participation by providing justifications for assignments. In contrast, authoritarian lecturers force the students to think or behave a certain way without considering their opinions. Furthermore, the environment (encouraging, well-prepared, pleasant) to which students are exposed to is vital for student motivation (Cayubit, 2022). When a quality learning management system (LMS) is made available to students in which they receive instant feedback and personalised interaction with peers and lecturers, it improves motivation (Ferrer, Ringer, Saville, A Parris, & Kashi, 2022; Zhang, 2025).

Some other prevalent teaching methods have the potential to diminish students' self-motivation. These aspects include the emotions students encounter like anxiety and stress (Liu, Shi, & Wang, 2022) and pressure to perform (Ryan & Deci, 2020). Lecturers are especially crucial in encouraging perseverance in learning environments by helping students feel that their studies have purpose (Bureau, Howard, Chong, & Guay, 2022). while peer interactions and support are a crucial contextual component to meet the fundamental psychological needs of students. Peer support, such as boosting one another and promoting a supportive learning environment, is highly linked to increased motivation and engagement (Wang, Sun, Wang, & Li, 2022).

It is essential to gain insight into how students' cognitive functions and motivation relate to their academic learning activities. Research emphasises the highly complex twenty-first-century abilities, such as problem-solving, logical reasoning, interpersonal skills, collaboration, innovation, creativity and self-directed learning, that must be developed to fully participate in deep learning (Mthethwa-Kunene et al., 2021; Reimers, 2021; Sølvik & Glenna, 2022). Moreover, these authors assert that EF offers the essential internal cognitive abilities required for students to effectively develop these skills. The main motivator to a deep learning approach requires higher-order mental skills like self-regulation, logical and analytical thinking and metacognition. Deep learning does not occur in isolation; rather, it happens when students employ their EFs and metacognitive skills to link difficult concepts, find deeper meaning and apply what they've learned to different situations (Reimers, 2021). Although a deep approach is essential, the ability to control oneself is ultimately what enables or prevents the use of these advanced cognitive abilities such as EF (Häsä, Rämö, & Yan, 2024).

In contrast to the deep approach, participating in simple, lower-order mental tasks that exclude extensive thinking is referred to as surface learning (Biggs & Tang, 2007). This strategy is motivated by the desire to do tasks as quickly as possible while still seeming to meet the task objectives. As a result, even when deeper, more complicated thinking is required to do the work successfully, students rely on easy solutions (Biggs & Tang, 2007). Students who lack self-regulatory skills; the capacity to organise, oversee, and assess their own learning, are inclined to stay at the surface level and depend on lecturers to provide them with precise instructions and guidance (De La Fuente, Sander, Kauffman, & Yilmaz Soylu, 2020). In addition, unstructured knowledge is usually the outcome of surface learning. This indicates that although students can recognise, label, or record isolated facts, they are unable to apply this knowledge to other situations (Biggs & Tang, 2007). As a result, students struggle to control their learning, track their development and perform cognitive tasks (EFs), like planning, managing time and decision making (De La Fuente et al., 2020; Reimers, 2021). This implies that students who successfully adopt a deep learning approach will be less likely to employ surface learning strategies.

The core components of EF – CF, WM and IC – predict deep learning and promotes self-regulated and goal-directed learning (Pinochet-Quiroz et al., 2022). Improved EF enables students to better organise, regulate and assess their learning, which helps them to cognitively grasp unfamiliar concepts as opposed to mindless memorisation (Dörrenbächer-Ulrich, Dilhuit, & Perels, 2024).

Nevertheless, students' adoption of a deep or surface learning approach is not entirely determined by EF. Rather, motivation, in particular intrinsic motivation as described by SDT (Ryan & Deci, 2020), guides students to use the required cognitive skills when engaging in learning (Stasolla et al., 2025). Harel-Gadassi (2022) states that using EFs, like creating objectives, self-regulation, self-assessment and planning, helps students feel more confident in their learning skills and boosts their level of autonomy. Excessive autonomy has an impact on intrinsic motivation because students who excel in their studies are more likely to be mentally invested in the learning activities, which in turn motivates them to be more committed (Harel-Gadassi, 2022).

Intrinsic motivation is thought of as the learning ideal that promotes real understanding, whereas extrinsic motivation is frequently seen as a temptation to embrace surface study techniques (Biggs & Tang, 2007). Extrinsic motivation and learning strategies have a complicated relationship. When it comes to surface learning, the impact is clear: students who are motivated by outside incentives typically put in less effort, rely on memorisation and concentrate solely on finishing the assignment (Acosta-Gonzaga & Ramirez-Arellano, 2021; Biggs, 1991; Biggs & Tang, 2007; Häsä et al., 2024). On the other hand, when students redirect that same external motivation into deliberate, advanced thinking to achieve a desirable goal, extrinsic motivation can also assist deep learning. Intrinsic and extrinsic motivational factors can also concurrently influence deep learning processes, each contributing uniquely to the student's engagement and comprehension in a complementary fashion (Ryan & Deci, 2017).

In contrast, amotivation is linked to negative emotions including disinterest, worry and scepticism, resulting in surface learning (Biggs & Tang, 2007). Amotivated students are more inclined to reduce their work if they do not believe the content to be inherently fascinating (Biggs, Kember, & Leung, 2001; Gozalo, León-del-Barco, & Mendo-Lázaro, 2020). A student who lacks the inherent motivation to interact with the material is said to be amotivated (also known as “unmotivated”), which usually leads to a surface rather than a deep approach to learning. When a student is unmotivated, it usually indicates that they lack belief that their work is worthwhile or they have no expectation to succeed (Biggs & Tang, 2007). Therefore, no motivated effort takes place if the likelihood of success is zero.

Students' cognitive abilities are influenced by their level of academic motivation (Berestova, Kolosov, Tsvetkova, & Grib, 2022), which implies that EF affords students the mental abilities they need to regulate their behaviour, but motivation acts as the driving force guiding students to use those abilities in their learning. Hence, this study aims to explore how academic motivation, as well as a deep and surface learning approach of DL student in HE within the South African setting associate with their EF skills. The following hypotheses are stated to explore this goal further.

An empirical investigation was conducted to survey undergraduate teaching students using a cross-sectional design. Given that the goal was to conduct a preliminary investigation of relationships between a set of variables that had not been previously reported, this approach was considered appropriate (Spector, 2019; Wang & Cheng, 2020).

The 25-item ESQ-R were used to assess students' EF skills related to the academic context (Strait et al., 2020). Five subscales, each representing a different EF skill, namely plan management, time management, organisation, emotional regulation and behavioural regulation, were assessed and recorded on a four-point Likert scale varying from 0 – never or rarely, 1 – sometimes, 2 – often and 4 – very often. Students responded based on how frequently they participate in behaviours related to each skill.

The 28-item college version of the Academic Motivation Scale (Vallerand et al., 1992) measured students' 1. Intrinsic motivation, 2. Extrinsic motivation and 3. Amotivation based on why they participate in certain behaviours. Responses were presented on a 7-point Likert scale ranging from one (does not correspond at all), 2–3 (corresponds a little), 4 (corresponds moderately), 5–6 (corresponds a lot) and seven (corresponds exactly). Students responded to the questions based on why they participate in certain behaviours.

The R-SPQ-2F (Biggs et al., 2001) was used to assess students' approach to learning. The scale consists of 20 items, which measure deep and surface learning. Deep- and surface learning were measured on a 5-point Likert scale varying from one “never or only rarely true of me”, two “sometimes true of me”, three “true of me about half the time”, four “frequently true of me” and five “always or almost always true of me”. Students responded to the questions based on their study experiences.

Several procedural practices were in place to avoid common method variance and bias: enhancing participant anonymity, increasing item reliability, isolating predictor and criterion variables, using different Likert scales and balancing question sequence (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003).

The applications, Mplus 9.0 (Muthén & Muthén, 2025) and SPSS v30 (IBM SPSS, 2025), were used for statistical analysis and factorial management in this study. Additionally, the maximum likelihood (MLM) estimator was employed with the covariance matrix as input (Muthén & Muthén, 2025). To assess the fit of the measurement and structural models, the root mean square error of approximation (RMSEA), the standardised root mean square residual (SRMR), the chi-square (χ²) indicator, the Tucker–Lewis Index (TLI) and the comparative fit index (CFI) were used. The SRMR values below 0.08 and CFI and TLI values higher than 0.9 (Byrne, 2010) were considered a suitable fit. Additionally, RMSEA values of 0.05 or less suggest a close or good fit, while values between 0.05 and 0.08 indicate an adequate fit (Byrne, 2010). The measurement models were compared using the Bayes Information Criterion (BIC) and the Akaike Information Criterion (AIC), which penalise for model complexities, with a greater match shown by lower values (Wang & Wang, 2020).

McDonald's omega coefficient (ω) was used to examine scale reliability. A cutoff value of 0.7 or greater was deemed ideal for interpreting composite reliability, although it is reasonable to anticipate results below 0.7 when working with psychological concepts (Field, 2018), and when only a few items measured the construct (Pallant, 2020). The Pearson product–moment correlations were incorporated to determine the associations among the variables, and a 95% level of statistical significance was chosen (p < 0.05). A correlation of 0.30 or higher indicates a moderate effect, but a correlation of more than 0.50 indicates a major impact (Field, 2018).

Teacher students associated with a private HE institution in South Africa were chosen to take part in an online survey. Data were collected from students between February and March 2024, before the integration of generative AI tools in HE. Every student was allowed to participate, and those who accepted the invitation were added to the research. Participants in the survey completed online questionnaires that were sent to them via a link. There were 917 respondents between the ages of 18–60. 32 males and 882 females participated in the study, and three participants preferred not to disclose their gender. Most participants reside in Gauteng (26.1%), and isiZulu (29%) and Setswana (25.1%) are the languages mostly spoken. Only 2.9% of participants indicated that their English language proficiency is below average. Participants (74.8%) have no teaching experience and only (26.8%) are in practice while studying. The participant characteristics are summarised in Table 1. Ethical approval to conduct the study from a recognised South African university was granted.

The initial measurement model was configured similarly to what theory proposes. However, in an initial run of the model, confirmatory factor analysis (CFA) indicated that the results were not ideal. To determine whether items of the different instruments (ESQ-R, AMS-C and R-SPQ-2F) loaded onto the factors that were intended according to theory, exploratory factor analysis (EFA) was utilised. The instrument items served as observed variables. Results indicated that the time dimension of EF, as well as two components of the motivation scale (introjected- and identified regulation) loaded across other constructs with no justification for retaining these items; consequently, they were omitted from the subsequent model. Items that loaded poorly <0.3: Wang & Wang (2020) onto their respective constructs were also removed before testing the final model.

The final measurement model was constructed as follows: (1) EF, a second-order construct with four first-order variables (planning, organisation, emotional regulation and behavioural regulation); (2) intrinsic motivation, a first-order construct measured by nine observed variables; (3) extrinsic motivation, a first-order construct measured by two observed variables; (4) amotivation, a first-order construct measured by four observed variables; (5) deep learning, a first-order construct by 10 observed variables; and (6) surface learning, a first-order construct with eight observed variables. All indicators considered in combination, the model demonstrated a suitable fit to the data (χ² = 2033.541*, df = 1056, p = 0.000, CFI = 0.91, TLI = 0.91, RMSEA = 0.03, SRMR = 0.05). The standardised regression coefficients all loaded positively and statistically significantly onto their respective constructs (p < 0.01). Table 2 presents the descriptive statistics, coefficient reliabilities and correlations of the instruments employed in this study.

The composite reliability of all the scales exceeded the threshold of 0.7, thereby indicating that the scales reliably captured the respective constructs.

Considering the maximum values that could be obtained for each of the various scales, above average mean values were recorded for EF, intrinsic- and extrinsic motivation as well as deep learning. The mean for surface learning was about average and the amotivation factor shows a desirable low mean.

The standard deviation for EF was relatively closer to 0 (0.43) indicating that participants' opinions on this construct were essentially more similar. However, larger variation in the participants' responses on the motivation and deep- and surface learning scales were detected. This suggests that individual responses varied greatly, representing a range of experiences and opinions, even while the mean values offer an insightful overall picture.

There is a statistically significant positive relationship between EF and intrinsic motivation (small effect); the association between EF and extrinsic motivation was negative (−0.063), but this result is not statistically significant. EF further showed a statistically significant positive relationship with deep learning (medium effect) and negative correlations approaching a large effect with both amotivation and surface learning.

The most robust positive correlations are observed between amotivation and surface learning (0.782, p = 0.00) and between deep learning and intrinsic motivation (0.602, p = 0.001) These associations suggest that students who are more intrinsically motivated are significantly more likely to use deep learning strategies while those who are amotivated are more inclined to use surface learning strategies.

Results furthermore revealed a statistically significant positive relationship between extrinsic and intrinsic motivation showing a large effect (0.502, p = 0.00). While both motivation types are evidently influential when it comes to learning, it is also interesting to note that while both intrinsic and extrinsic motivation are positively associated with both deep- and surface learning strategies, intrinsic motivation is more strongly correlated with deep learning while extrinsic motivation is more strongly correlated with surface learning. Finally, there is a statistically significant positive relationship between deep and surface learning showing a small effect. As expected, the relationship between amotivation and intrinsic motivation was not significant.

The measurement model served as the basis for testing the structural model. The results revealed an acceptable fit of the data (χ² = 2033.541*, df = 1056, p = 0.000, CFI = 0.91, TLI = 0.91, RMSEA = 0.03, SRMR = 0.05).

EF reflected a negative association with surface learning in the structural model, but the path coefficient of EF (−0.06) is not statistically significant (p > 0.05). Therefore, H1b is accepted.

Amotivation (coefficient = 0.733) has the strongest and most significant positive (p < 0.001) effect on surface learning. This finding implies that those who are more amotivated are more likely to employ surface learning techniques. In addition, extrinsic motivation also has a significant positive effect (0.2, p = 0.00) on surface learning, resulting in extrinsic motivation to encourage one to engage in surface learning. H2b and c is confirmed. Although the direction of the relationship between intrinsic motivation and surface learning was negative, the result was not statistically significant (−0.043, p > 0.05). H2a is not accepted. These findings suggest that individuals with higher levels of intrinsic motivation will be less likely to employ surface learning motives and strategies.

Three factors, namely, EF, extrinsic- and intrinsic motivation, are statistically significant predictors of deep learning strategies. Intrinsic motivation shows the strongest effect (coefficient = 0.468, p = 0.00) followed by EF (0.298, p = 0.00) with the second largest effect. Extrinsic motivation also played a role, but this effect was rather small (0.194, p = 0.00). The fourth factor, amotivation, played no role in deep learning (p > 0.05). These findings suggest that increases in EF as well as two types of motivation (intrinsic and extrinsic) can be beneficial to support a deep learning approach. H1a and H3a and b is accepted; H3c cannot be accepted.

The current study advances our knowledge of the connections between teacher students' academic motivation, EF and deep and surface learning approaches. The aim was to investigate how these aspects associate with students' EF skills of DL students in HE within the South African setting. The results offer important insights into how cognition and motivation simultaneously impact learning processes in a DL environment.

In this study, EFA and CFA confirmed that only four out of the proposed five-factors of the ESQ-R subscales were reliably measured. The factors include planning, organisation, emotional regulation and behavioural regulation. Due to cross-loading, the three items measuring time could not be reproduced as a separate sub-component of EF and was therefore excluded from the analysis. This finding contrasts with the initial development of the ESQ-R scale by Strait et al. (2020), which proposed time as a separate sub-dimension. However, the exclusion of the time factor in the current study corroborates with a recent Mexican study conducted with students from 12 universities, which developed and validated a self-report EF scale suitable for HE students (Zamora-Lugo et al., 2025). Their study analysis revealed that the time dimension could also not be measured separately but instead formed part of a larger latent variable and was therefore integrated with the planning dimension (Zamora-Lugo et al., 2025). The close association between time management and planning is also in line with research by Knouse, Feldman and Blevins (2014).

Additionally, in the current research, the sample was drawn from a non-Western cultural context. The participants represented a range of rural, semi-urban and urban settings across South Africa, which may have influenced the findings. As highlighted by Dörrenbächer-Ulrich et al. (2024) and Ramos-Galarza et al. (2023), it is possible that time is experienced and perceived differently across cultures and contexts. This validates the idea that EFs are rooted in culture and that societal norms regarding time and accountability influence students' perspectives of time (Zamora-Lugo et al., 2025). The findings indicate that further research is required to better understand the time variable within the SA HE context.

In relation to the hypotheses presented, the findings offered supportive empirical proof. The findings confirmed that EF had a statistically significant positive effect on deep learning (H1a). This implies that students who possess greater EF abilities like WM, CF and IC are better equipped to adopt a deeper approach to learning. This finding provides further support for previous research suggesting that goal-directed behaviour and cognitive control may promote meaningful interaction with the subject matter (Pinochet-Quiroz et al., 2022). This is also consistent with research that associates EF with self-directed learning serving as the foundation for academic success (Dörrenbächer-Ulrich et al., 2024; Harel-Gadassi, 2022).

In contrast, a negative association between EF and surface learning was detected; however, the directional effect (H1b) of EF was not statistically significant. The negative correlational result suggests that EFs may inhibit inappropriate learning strategies and promote more self-regulated and focused learning. Broadly speaking, this result resonates with research by La Lopa and Hollich (2014), which claimed that higher level of EF is likely to a decrease dependence on surface learning techniques, which are usually linked to rote memorisation and low cognitive effort. In addition, students not remembering information over time results from lecturers and students failing to consider the rigorous capacity limits of WM, which leads to a dependence on rote memorisation and simple surface learning techniques (La Lopa & Hollich, 2014). Therefore, one can note that higher levels of EFs equip students with the skills to organise, transform and integrate knowledge instead of merely storing it.

The correlational relation result furthermore corroborates with the results of a study which concludes that students who apply deep learning techniques exhibit greater knowledge of difficult topics, while those who use surface learning techniques lack analytical abilities and only depend on memorising the content (Paleenud et al., 2024). It should however be kept in mind that although the structural model result also indicated a negative association between EF and surface learning, the directional effect was not statistically significant and should therefore be further examined in future studies.

The second main hypothesis examined how different types of motivation affected surface learning strategies. Both extrinsic and amotivation showed statistically significant positive influences on surface learning, which is consistent with Hypotheses 2b and 2c. When students are extrinsically motivated, they are more likely to rely on surface learning. This finding recalls on research conducted by (Biggs, 1991) emphasising that extrinsic motivation and surface leaning are directly related to each other. According to an existing study, it is possible that surface learning can be a result of factors such as anxiety, low task interest and extrinsic motivation (Biggs & Tang, 2007). In addition, students may not have the socioeconomic motivation necessary to participate in deep learning, especially in rural communities that lack academic expectations. This could lead to reduced motivation and a greater tendency towards surface techniques (Fu & Liu, 2024a, b).

Moreover, in this study, the largest predictor of surface learning is amotivation, suggesting that students who feel unmotivated or believe that learning tasks are not worthwhile are more likely to use surface learning methods. This finding corroborates with existing studies which suggest that the degree to which students may choose low-effort learning techniques may be due to discouragement, an absence of value, or doubt about their competence and a fear of failure, and these indicators of amotivation may inevitably result in surface – rather than deep learning (Biggs & Tang, 2007; De La Fuente et al., 2020; Hulreski, Syatriana, & Ardiana, 2020). Furthermore, literature notes a crucial connection between amotivation and surface learning with the topic of self-regulation. With a lack of self-regulation, students will adopt a surface approach to learning since they find it difficult to control their learning (De La Fuente et al., 2020).

As anticipated in Hypothesis 2a, intrinsic motivation had a negative impact on surface learning; however, this effect was not statistically significant. The lack of statistical significance demonstrates that intrinsic motivation alone may not consistently prevent surface learning behaviours. Subsequently, Hypothesis 2a was not accepted. This result emphasises how complex motivation is as it comprises multiple interrelated components like values, beliefs, views, opinions and behaviour amongst others (Abdolrezapour et al., 2023).

According to findings of the third main hypothesis, deep learning would benefit from intrinsic and extrinsic motivation, but amotivation would have the opposite effect. H3 emphasises the significant role of motivation in enhancing deep learning, which is mostly aligned with the findings of this study. One of the strongest effects found in the study was the strong and statistically significant positive relationship between intrinsic motivation and deep learning (H3a). When students are driven by curiosity and personal meaning, they are more likely to invest in cognitive effort necessary for understanding concepts and critical thinking, resulting in deep learning. This result is in line with previous research by Pinochet-Quiroz et al. (2022) which showed that DL inherently requires higher levels of self-regulation and autonomous functioning.

Additionally, there was a statistically significant, but weaker, positive effect of extrinsic motivation on deep learning (H3b). Together, the findings of H3a and b are in line with the SDT framework which considers extrinsic and intrinsic motivation as two aspects on a continuum of self-regulation, rather than as two independent categories (Gareau, Gaudreau, & Boileau, 2019; Ryan & Deci, 2017). According to Harel-Gadassi (2022), extrinsic factors, such as rewards, may be the initial stimulus to perform a certain activity but as learning advances, the extrinsic factors can increasingly be internalised and converted into intrinsic motivation. Moving along this continuum may lead students to become increasingly autonomously motivated to participate in learning tasks for their own intrinsic fulfilment and worth, which is ultimately linked to better academic results, increased well-being and lower levels of stress.

While the study hypothesised that amotivation would have a negative effect on deep learning, this relation was not supported, indicating no significant effect (H3c). Consequently, amotivated students will likely adopt surface learning strategies, while intrinsically motivated students are inclined to engage in deep learning techniques. Prior research relates amotivation with deep learning, mainly because it keeps students from developing the self-control and cognitive skills required for meaningful learning (Berestova et al., 2022). The result of the current study does not support this finding.

The findings imply that an improvement in EF, as well as intrinsic- and extrinsic motivation may benefit students and encourage them to adopt a deep learning approach. Collectively, these results highlight the intricate interplay between cognitive and motivational processes in influencing students' learning approaches. Although EF constitutes the cognitive underpinnings for deep learning, motivation, especially intrinsic motivation and some forms of extrinsic motivation are substantial factors influencing how well these cognitive abilities are used in students' learning. This result corroborates with previous research which found that intrinsic and identified extrinsic motivation relate to deep learning, whereas external regulation and amotivation are associated with surface learning (De La Fuente et al., 2020). In conclusion, the findings confirmed the hypothesis that intrinsic and extrinsic motivation, as well as EF, have a statistically significant positive effect on deep learning, and no noticeable correlations exist between EF and surface learning and EF and amotivation.

The results of this study have many significant implications for research, methodology and teaching in distance HE. Early identification and support could be beneficial for both students and the institution (Harel-Gadassi, 2022; Trolian & Jach, 2020) in achieving academic success. It is therefore recommended for institutions to implement online learning environments that offer strong support for strengthening EF skills and cultivating more autonomous learning strategies. It is essential to identify students with lower EFs when admitting them to the institution, especially first-year students. This is crucial since these students are at high risk of failure in the self-regulated DL environment (Harel-Gadassi, 2022). After identification, adequate guidance should be provided, which involves teaching them certain skills and enhancing their feelings of intrinsic motivation and self-confidence (Harel-Gadassi, 2022). In addition, continuing evaluation of students' self-report EF progress longitudinally may offer vital insight to institutions which can be used to support students at risk continuously.

Furthermore, effective student support involves various cognitive and meta-cognitive strategies to enhance EF and self-regulated learning. Activities comprising problem-based learning (PBL), in which students are expected to solve real-world problems and apply it, is an effective way to build twenty-first century skills and enhance students' understanding of a particular concept (Biggs & Tang, 2007; Martinez, 2022). Clear study guidelines, reasonable goal setting, regular formative evaluation and concrete resources such as diagrams, concept maps and digital instruction during problem-solving (Ertmer, Schlosser, Clase, & Adedokun, 2014) are all effective ways to enhance autonomy in DL (De La Fuente et al., 2020). It is crucial to equip students with the cognitive tools such as creating questions while studying, summarising information using visual aids and actively engaging with the material to help them recall difficult content (Corcoran and O’Flaherty, 2017).

In this advanced digital world, the provision of innovative technological tools could ensure structure and within the DL context. Mobile AI chatbots act as a supportive digital assistant that enhance students' EFs: CF, WM, problem-solving skills, metacognitive skills, planning, creativity and self-regulation by providing timely feedback in a supportive and organised environment (Dias, Avila, da Costa, Cardoso, & Fonseca, 2022; Pergantis et al., 2025; Zhang, 2025). Moreover, implementing e-portfolios might guide students to engage in continuous self-monitoring and critical reflection on their learning and development (Reimers, 2021). According to Aderibigbe (2021), Community of Practice (COP) groups and processes can be established to regulate and motivate students to participate in planned and organised online discussions. This method enables lecturers to supervise and direct conversations, identify learning gaps and provide students with focused support. Mentoring students in this way may improve their academic self-efficacy, address EF impairments and increase self-confidence. Hence, having access to a high-quality LMS that allows students to connect with peers and lecturers, increases motivation (Ferrer et al., 2022). Implementing the latter will ensure that institutions focus on the process of learning, rather than the product, resulting in enhanced academic motivation.

It is recommended that institutions implement an integrated learning and teaching strategy that increases both intrinsic and extrinsic motivation while also enhancing students' EFs. Although intrinsic motivation is the primary aim, using both intrinsic and extrinsic rewards could be useful to encourage unmotivated students (Hidi & Harackiewicz, 2000). When intrinsic motivation is low, lecturers can use properly designed extrinsic incentives together with meaningful tasks to encourage greater interest (Hidi & Harackiewicz, 2000). Additional studies should explore the long-term effects of focused interventions that incorporate motivation and EF training on students' learning strategies. Increased understanding of the causal relationships between EF, motivation, deep learning and academic achievement may be possible through longitudinal and experimental research in various contexts. Additionally, employing qualitative research to collect detailed insights from students' perspectives could produce insightful data.

In the South African DL environment, where students study without face-to-face contact, strong EF skills seem crucial for managing learning independently and applying knowledge rather than engaging in rote learning. Since many students enter HE with diverse educational backgrounds and little exposure to self-directed learning, this study raises the possibility that DL may unknowingly increase existing gaps between students with higher and lower EF skills (Dörrenbächer-Ulrich et al., 2024; Harel-Gadassi, 2022). Expectedly, DL increases HE access but also places pressure on students' EFs and motivation. Hence, without lecturer and institutional support, it risks study dropout among students with lower EFs or higher amotivation (De La Fuente et al., 2020). In summary, in the South African HE context, DL should involve quality EF and motivation support for students, focused support systems and curricula must be implemented, and finally it is essential to encourage deep learning across support, assessment and learning programs.

This study's findings could assist HE institutions to create interventions and develop frameworks that support cognitive and motivational aspects of learning. The findings can be used as the foundation for initiatives intended for short courses on EF enhancement, self-regulated learning (e.g. planning, goal setting) and motivation-intensive training to encourage students to become self-sufficient, confident, autonomous and competent in their learning. By applying these strategies in practice, HEI may guarantee that DL is inclusive and successful for students who face cognitive and motivational difficulties, as well as for students with strong self-regulation abilities.

The limitations of this research is acknowledged. Initially, participating students in the study was predominantly female (96.2%), restricting the findings applicable to varying gender populations. Therefore, the findings should not be generalised to other undergraduate teacher–student groups in South Africa. Similarly, no separate analysis pertaining to factors like age, gender and culture was performed; therefore, more research is recommended. Future studies should aim to incorporate equitable gender balanced samples from several HE institutions to increase cultural significance.

Another limitation pertains to the use of self-report questionnaires to collect data, which may have resulted in inaccurate results in some instances, even after every effort was made to minimise response bias. It's possible that inflated responses resulted from students' differing perceptions of the constructs. Nonetheless, as self-report instruments continue to be the most successful way to evaluate students' views, thoughts and experiences, applying them was inevitable. It is suggested that future studies include mixed-methods research designs, integrating qualitative strategies to allow for a more in-depth understanding of students' reflections and perspectives on EF, motivation and learning.

It is vital to note that data were collected before the extensive incorporation of generative AI tools. Therefore, the current study did not account for the potential impact of AI-supported methods of learning on student participation, motivation and learning approaches. Future studies should specifically look at how deep- and surface learning processes in digital learning settings might be affected by AI-driven assistance.

Employing a cross-sectional study methodology, which precludes measuring factors that change with time, is the third limitation. As a result, the relationships between the different factors were interpreted and not verified. To evaluate long-term outcomes and the relationships between EF, deep and surface learning and various forms of motivation: intrinsic, extrinsic and amotivation, longitudinal studies are therefore recommended for further research. Considering the study's cross-sectional design, these relationships can only be conclusively demonstrated through recurring research.

This study explored the associations between EF, academic motivation and learning approaches among student teachers in a DL setting. Recognising the relationship between EF, a deep learning approach and academic motivation are essential to helping students succeed in a demanding context like HE. The findings reveal that EF and intrinsic motivation are significantly linked to deep learning, while amotivation mainly predicts surface learning approaches. This study contributes to current theoretical perspectives claiming that motivational quality and cognitive control play a significant role in how students deal with learning in HE. The results offer empirical data regarding the function of EF and motivation in students' learning approaches within a digitally facilitated learning context. Subsequent studies can expand on such results by employing longitudinal or mixed-methods methodologies to further investigate how these associations evolve over time and in various learning environments.

The study was conceptualised, reviewed and amended with input from LP and MH. MH carried out the statistical analysis, and both authors accepted the submitted version of the article and contributed to it.

Ethical clearance was obtained from the Basic and Social Sciences Research Committee (BaSSREC) of the Northwest University in Potchefstroom, South Africa (number: NWU-01046–23-A7).

This manuscript's expressions are entirely the authors’ own and do not necessarily reflect those of the publisher, editors, reviewers or related institutions.