Working Memory (WM) is a cognitive process responsible for processing and manipulating information while multitasking with other cognitive operations (Baddeley, 1992). WM is essential for directing and sustaining goal-related information despite the appearance of distracting stimuli (Schmeichel et al., 2008; Hofmann et al., 2012). Previous studies have found a positive relationship between WM and self-regulation skills, including attention (Diamond and Ling, 2020), inhibitory ability (Engle, 2002), and emotion regulation (Schmeichel et al., 2008; Xiu et al., 2018). People who score well on WM ability have demonstrated superior performance in focusing on tasks. In contrast, individuals with learning difficulties, self-regulation and attention deficits may have impaired WM and find it hard to inhibit inappropriate behaviour and effectively manipulate and combine goal-relevant information leading to variability in learning (Westerberg et al., 2004; Martinussen et al., 2005; Grégoire et al., 2012; Van Snellenberg et al., 2016; Xiu et al., 2018). Therefore, there is a robust theoretical and empirical connection between WM and academic performance.

WM training has been implemented to enhance academic performance, given its strong association with cognitive functioning (Unsworth and Engle, 2007). Research has demonstrated significant improvements in both trained and untrained tasks after WM training, a phenomenon known as “transfer.” Near transfer refers to gains in untrained but similar tasks, such as WM updating tasks (Borella et al., 2013), while far transfer refers to gains in untrained tasks that share similar cognitive processes with the training tasks, such as intelligence and verbal ability (Karbach and Verhaeghen, 2014; von Bastian and Oberauer, 2014).

Furthermore, WM training has been shown to improve functional connectivity between the prefrontal cortex and posterior parietal cortex, as well as boost dopamine levels, suggesting that it refines the functional WM network and could lead to wider cognitive gains (McNab et al., 2009; Jones et al., 2022). However, the retention of such gains and the duration of data collection are often under-specified in the literature. More consistent reporting is needed to gain a better understanding of the effect.

Several theoretical mechanisms have been proposed to explain the benefits of WM training, such as neuroplasticity and process-specific theory (Klingberg, 2009; Gathercole et al., 2016). However, these mechanisms do not fully account for the transferable effects of WM training. Söderqvist and Bergman Nutley (2015) proposed a theoretical framework comprising learning and performance routes that considers how WM training improves the ability to grasp information and engage in academic tasks, leading to the establishment of routines that can be transferred to other cognitive tasks.

Moreover, WM training can benefit brain networks that support not only cognitive but also emotional functioning (Pessoa, 2013). Emotion regulation affects learning experiences and academic success, is intertwined with many cognitive processes and provides motivation and engagement with the environment (Sutton and Wheatley, 2003; Graziano et al., 2007; Fried, 2011). WM training can strengthen and extend cognitive routines, enhance inhibitory and regulatory abilities, and improves coping with daily cognitive tasks and negative emotional experiences in the frontoparietal network (Schweizer et al., 2013).

Despite a considerable amount of evidence supporting the efficacy of working memory (WM) training for learning, there is still a gap in our understanding of its long-term impact in educational settings, as well as how it can be effectively delivered in the classroom and what factors influence its effectiveness in practice. The purpose of this mini review is two-fold: first, to clarify the theoretical foundations of WM training and transfer, and second, to investigate the conditions and variations in the implementation of WM training in educational contexts, including how educators perceive and deliver this training, and ultimately, how it affects student learning outcomes. Additionally, this review highlights the need for a collaborative partnership between researchers and educators to support the effective implementation of WM training in schools.

This mini narrative review therefore aimed to synthesise and summarise the key findings and challenges in the literature related to WM training in the classroom context. The search was conducted in online databases, including Web of Science, PsycINFO, PubMed, Education Resources Information Center (ERIC), and Google Scholar. The author also used snowballing methods, which included backward and ancestry techniques, to identify relevant articles by searching through their bibliographies and reference sections. Additionally, citation-tracking methods (e.g., bibliographic coupling) were utilised to identify forward-looking research.

The search strategy included a combination of relevant keywords, such as “working memory training,” “implementation,” “intervention,” “educational practice,” “classroom,” “school context,” “challenges,” “dilemmas,” “applicability,” and “transferability.” Boolean operators were used to refine the search, including AND, OR, and NOT.

Inclusion criteria for studies were: (1) both secondary research and empirical research studies on implementing WM training in educational settings (e.g., schools, colleges, universities) or discussing the implementation of WM training for educational practice, (2) published in peer-reviewed journals across native English-speaking countries, (3) focused on children in compulsory education from ages 5–16 years, post-16 education or students in higher education institutions, and (4) available in full-text format.

To ensure the quality and relevance of the studies included in this review, the author followed a rigorous screening process. The author conducted a preliminary search and screened articles based on their relevance to the topic and that they address the review research questions and met the inclusion criteria. The author excluded studies that did not meet the inclusion criteria. After the initial screening process, the author reviewed the full-text versions of the remaining articles to ensure they met the inclusion criteria. The author also searched the reference lists of these articles for additional relevant studies that were missed in the initial search. Finally, the author extracted relevant findings from the included articles to synthesise and summarise the key findings and challenges in the literature.

WM training has the potential to produce near transfer effects, but far transfer effects are less common. WM training has been found to be beneficial for students in areas related to language, such as reading comprehension and listening comprehension (Artuso et al., 2019), as well as in mathematical abilities such as numerical capacities, reasoning, and arithmetic skills (Holmes et al., 2009; Kuhn and Holling, 2014; Söderqvist and Bergman Nutley, 2015), emotion regulation (Xu et al., 2021), attention level (Volckaert and Noël, 2015), and intelligence (Jaeggi et al., 2008). However, a similar volume of research evidence has shown limited or no improvement after training. For example, Dunning et al. (2013) found that while training improved spatial short-term memory, verbal WM, and visuo-spatial WM, it had no effect on verbal short-term memory (Verhagen and Leseman, 2016). Thorell et al. (2009) similarly noted that training failed to enhance attention and general fluid intelligence, with the transfer of training effects to WM capacity being limited to visuo-spatial tasks. This variability in findings is also reflected in several published reviews. Sala and Gobet’s (2017) meta-analyses of experimental studies showed a robust near-transfer effect on relevant memory tasks, but only a small to moderate far transfer effect on fluid intelligence, cognitive control, mathematics, and literacy. Melby-Lervåg and Hulme (2013) and Melby-Lervåg et al. (2016) conducted meta-analyses on computerised WM training in typically developing individuals, and their findings suggest that the evidence for near transfer is mixed and the claims for far transfer effects are poorly supported. They found that short-term training effects only supported near transfer effects when they were specific to the trained task and did not generalize to other areas, such as word decoding, verbal ability, and arithmetic outcomes. As Dahlin et al. (2008) suggested, post-test gains would only be observed when the measurement tasks share the same cognitive function as the training tasks.

One important factor to consider is the duration for which the effects of WM training persist. Although there is no clear consensus on what constitutes short-term or long-term retention, this literature review will provide a brief classification based on evidence from empirical studies.

Although theories have attempted to explain the effects of WM training, the inconsistency in achieving transfer effects remains unclear (Jaeggi et al., 2012). A lack of investigation into how experimental and individual characteristics moderate training outcomes makes it difficult to determine whether different components of WM training follow the same mechanism and require further investigation. According to the Context Mechanism Outcome Configuration (CMOc) framework, the mechanism cannot be separated from the context and the individual’s reasoning in response to the program (Dalkin et al., 2015). As per Pawson and Tilley’s (1997) realist evaluation, a mechanism is not merely a theory that refers to program activities, but an interplay process that explains how a specific outcome was achieved. Therefore, given the diverse individual- and implementation-related characteristics across WM training studies, the current mechanisms are less informative for real-world application.

To better understand the underlying mechanisms that drive effective training, researchers and practitioners need to refine theoretical frameworks and create more robust experimental approaches. Factors such as different training methods, experimental settings, and individual differences of both trainees and program implementers should be considered as they contribute to the theoretical mechanisms that underlie effective training programs, which may or may not have practical value for educators.

WM studies often overlook the roles of researchers and educators, as well as their interaction and collaboration, and the external support and resources obtained (MacMahon et al., 2022). The traditional school-based implementation follows a unidirectional partnership, where researchers disseminate and interpret findings, and educators adapt the designs to fit their educational purposes. However, this approach fails to address the research-practice gap and ecological validity of research evidence (Daniel and Pollmann, 2012; Palinkas and Soydan, 2012; Joram et al., 2020; MacMahon et al., 2022). Van Atteveldt et al. (2018) proposed a bidirectional collaborative approach that involves educators in the design and implementation of research to facilitate research translation. However, educators and researchers have different epistemological and philosophical beliefs, and it is unclear how their perspectives may influence the decision-making process and implementation outcomes. Future research must explore how to incorporate diverse perspectives into all stages of the implementation process.

In terms of implementer-related variables, the impact of teachers’ motivation and self-efficacy on implementation outcomes has been considered (Durlak and DuPre, 2008). Knowledge brokers have been proposed to help translate research into other contexts in an accessible way (Malin and Brown, 2019; Cooper et al., 2020). However, concerns exist about the necessary financial or time expenses and coordination for such support. Therefore, those involved in implementation should work collaboratively to establish contextually compatible and robust translation, taking into account various factors.

Randomized controlled trials (RCTs) are highly valued for evaluating interventions that can inform educational decision-making (Burnett and Coldwell, 2021). However, implementing RCTs in schools can be challenging due to high attrition rates, cost-effectiveness, and potential overestimation of effect sizes due to insufficient sample sizes (Cheung and Slavin, 2016; Dawson et al., 2018). Researchers have suggested recording participant information in detail to address low response rates or high dropout rates, but such solutions may not always be contextually appropriate and require further empirical studies. Alternative methods, such as quasi-experimental designs, have been proposed but have been criticized for failing to clarify the mechanisms underlying treatment effects (Gopalan et al., 2020).

Another consideration regarding WM training within education is the issue of measurement tasks. The training effect is often calculated by by comparing performance at pre-test and post-test, but this may not sufficiently demonstrate the extent to which experimental manipulation modifies performance compared to a baseline level (Goodhew et al., 2020). Moreover, the measurement tasks adopted across studies are diverse. For example, Sánchez-Pérez et al. (2018) used the Spanish version of the Woodcock-Johnson III (WJ-III) Achievement battery to measure mathematical skills, while Oka et al.’s (2021) study used mental arithmetic, and Studer-Luethi et al. (2016) used a battery of phonological encoding, recoding, and reading comprehension to evaluate reading ability. Such varied assessments may account for inconsistencies in findings regarding the nature and timing of WM transfer effects. Furthermore, the relationship between the measurement tasks used in WM studies and academic tests in schools is not yet clear. Although the purpose of WM training is to generalize transfer from a training program to academic performance in school, experimental studies often overlook this connection.

Implementation frameworks may serve as a bridge for successful WM training implementation in school contexts. For instance, Meyers et al. (2012) Quality Implementation Framework outlines critical steps and phases throughout the implementation process, including decisions on preparations, structuring support and resources, and reflection or evaluation. Durlak and Dupre’s (2008) ecological framework, which looks at factors at the community, provider, innovation, and delivery system levels and provides a clear vision for both researchers and implementers to consider relevant elements involved in the process. However, empirical research is needed to examine how these implementation frameworks are integrated, how the factors in these frameworks contribute to implementation outcomes in educational contexts and how frameworks can be adapted to examine different schools and relate to a range of practitioner-level characteristics and knowledge.

For successful implementation of WM training in education, prospective studies should incorporate an evaluation of process and a framework for assessing the quality of study design to achieve educational purposes.

Realist social theory underlying the CMOc framework highlights the interconnections between individuals and society. By adopting realist social theory, researchers can critique “evidence-based” claims in a more nuanced manner (Archer, 1995; Clegg, 2005). Realist social theory acknowledges that the relationship between mechanisms and outcomes is not always straightforward, and that each project implementation may activate or reconfigure pre-existing environmental mechanisms in unique ways, resulting in a variety of potential outcomes. By evaluating these outcomes, researchers can identify when and for whom the project is effective (Pawson and Manzano-Santaella, 2012). This approach has proven valuable in studying school intervention programs (Warren et al., 2020).

The implementation of research evidence is influenced by contextual attributes (Tabak et al., 2018), such as ecological validity, which refers to the extent to which findings from a study can be applied to real-life settings (Andrade, 2018). However, the transfer of psychological research from controlled laboratory environments to educational settings can present challenges. Schmuckler (2001) proposed that ecological validity can be assessed based on three dimensions: (1) the nature of stimuli, (2) the nature of task, behaviour, or response, and (3) the nature of the research context. Each dimension could be evaluated based on the degree of artificiality–naturality (Hoc, 2001; Schmuckler, 2001) and simplicity–complexity (Peelen and Kastner, 2014; Lappi, 2015). To improve the generalizability of research evidence, researchers can explore factors and strategies that promote ecological validity and facilitate transfer to real-world settings.

Developing an effective evidence-based approach requires a collaborative implementation process, where educators’ team building, collaboration, and leadership are critical factors (Jones et al., 2022). Ford and Sutton (2009) indicated that effective collaboration between researchers and educators is essential to enhance students’ learning experiences. To promote teacher participation and collaboration with researchers, it is important to understand the factors that influence the establishment of a bidirectional partnership from the teachers’ perspective (Jones et al., 2022). A key issue is how teachers with varying research skills can assume an equal and proactive collaborator role. Teachers need to understand how the research being conducted can support their teaching and professional development to engage in research (Muhonen, 2014). Therefore, offering professional development opportunities to acquire research knowledge and skills, as well as access to educational resources and tools, is crucial (Kuntz, 2013). School leaders also have an important role in establishing a culture of research and continuous learning to support teachers’ research aspirations and long-term engagement with research (Cantalini-Williams et al., 2016; Cramp and Khan, 2019).

Such bidirectional relationships are informed by the interaction between the teaching team and researchers and extends to school administrators, parents, and stakeholders, whose value and investment can strengthen the implementation outcomes. Only when the agenda of teacher participation in research is given attention and prioritised by external parties can teachers’ self-efficacy and agency in research be enhanced. As Boyer (1991) notes, research should conceptualise the interaction between research and teaching to promote their integration and achieve a mutually reinforcing and shaping effect.