Working memory (WM) refers to the ability to actively maintain and process information needed to complete complex tasks such as comprehension, learning, and reasoning. Recent studies have examined the efficacy of computerized working memory training (WMT) in improving cognitive functions in general and WM in particular, with mixed results. Thus, to what extent can WMT produce near and far transfer effects to cognitive function is currently unclear. This study investigated the transfer effects of a computerized WMT for preschool children and also examined the possible neural correlates using the event-related potential (ERP) technique. A total of 50 Chinese preschoolers (64.44 ± 7.76 months old; 24 girls) received 4-week training during school hours. Compared with those in the active control group, children in the training group showed better gains in behavioral performance in the WM task and significantly more changes in ERP markers of the WM and inhibitory control tasks (near transfer effect). However, no evidence was found for transfer to fluid intelligence (far transfer effect). These findings suggest that WMT is capable of enhancing cognitive functioning in preschool children, and as such this work has important implications for educational practice and it may help to design and refine cognitive interventions for typically developing children and those with WM problems or other cognitive deficits (e.g., children with attention-deficit/hyperactivity disorder).
Working memory (WM) refers to a limited capacity system allowing the temporary storage and manipulation of information necessary for complex tasks such as comprehension, learning, and reasoning (Baddeley, 1992, Baddeley, 2012). WM is considered to be at the core of human cognitive abilities and plays a crucial role in learning, reasoning, problem-solving, and intellectual activity (Baddeley, 1992). Working memory capacity (WMC) is a positive correlate with other cognitive abilities, especially inhibitory control (Redick et al., 2011), and it is closely related to nonverbal intelligence (also called fluid intelligence; Engel de Abreu et al., 2010) as well as verbal intelligence, such as mathematical, reading and writing skills (e.g., Sedek et al., 2016). In recent years, a growing number of studies have examined whether repeated training in WM paradigms is capable of improving WMC and, if it does, whether such a training effect can generalize or transfer to other untrained tasks, including tasks of inhibitory control, fluid intelligence, and academic achievement (e.g., Jaeggi et al., 2008, Swanson and McMurran, 2018, Wang et al., 2022, Zhang et al., 2018, Zhao et al., 2013). The transfer effects on tasks closely resembling those trained are termed near transfer effects, whereas those on other cognitive tasks quite different from those trained are called far transfer effects (Melby-Lervåg et al., 2016). Despite the increasing interest in working memory training (WMT), different studies have reported contradictory findings concerning the degree of improved WM performance, the magnitude of far and near transfer effects, and specific structural and functional neural changes.
Preschool is a significant developmental period for children’s cognitive function, social communication, and emotional and brain maturity. WM begins to develop most rapidly in the preschool period (e.g., Cheng and Kibbe, 2022, Ross-Sheehy et al., 2021), and these changes occur in parallel with structural and functional changes in the nervous system involving the prefrontal cortex (Diamond, 2013). The development of WM in the preschool period is such an important event for children that their WMC often closely relates to their mathematic, reading, and writing achievements in the future (Sedek et al., 2016). For example, WMC in the preschool period is found to predict school-age mathematic and reading achievements (Shipstead et al., 2012), academic achievements (Snyder & Munakata, 2010), and fluid intelligence (Engel de Abreu et al., 2010). Given that the preschool period witnesses a high degree of brain plasticity and cognitive plasticity, cognitive training might have a better and stronger effect during this period than other periods of a person’s life (Wass et al., 2012). Therefore, understanding the transfer effects of WMT from a developmental perspective may help to understand its efficacy in a stage-specific manner. Research on WMT in preschool children can also help to design better cognitive training and intervention programs to help the behavioral development of children, especially those with mild cognitive impairments (e.g., autism, attention-deficit/hyperactivity disorder [ADHD]), develop better cognitive abilities. In addition, WMT is shown to strongly influence other components of executive function, and higher executive functions have a long-term impact on children’s lifelong achievement, health, wealth, and quality of life (Diamond, 2013); thus, WMT in the preschool period has clinical significance.
Despite its significance, research on computerized WMT for preschool children is rather limited, and evidence of far transfer is scarce. Thorell et al. (2009) adopted the Cogmed JM program to train WM of the visuospatial sketchpad in 4- and 5-year-old preschool children for 5 weeks. They found that this training brought significant improvements in untrained WM tasks but had no effect on inhibitory control, as assessed by the day–night Stroop and go/no-go tasks, or on fluid intelligence. Bergman Nutley et al. (2011) also chose the Cogmed JM program to train 4-year-old children but failed to observe any significant improvement in fluid intelligence scores in the WMT group. Zhang et al. (2019) conducted 20 sessions of WMT using the WM animal span task with preschool children. They also did not find any improvement in inhibitory control measures of the AX–CPT (the AX version of continuous performance task) and post-test fluid intelligence scores. Nevertheless, Peng et al. (2017) used the n-back task to train preschool children, and their results illustrated that the training group showed higher fluid intelligence scores in the post-test and in the 6- and 12-month follow-up tests.
Based on the above analysis, we thought that there may be two reasons for the discrepancy in the reported findings. First, the training programs target different components of WM across different studies. The Cogmed program (https://www.cogmed.com), designed and developed by Klingberg and colleagues (Klingberg et al., 2005), is one of the most widely used WMT programs. Cogmed JM is the version specifically tailored for 4- to 6-year-olds. The Cogmed JM program contains short-term memory and WM tasks; the WM tasks target both the storage and manipulation components of visuospatial WM (Yu et al., 2015). The WM animal span task also contains both storage and processing components that emphasize the maintenance function in the face of distraction (Oberauer et al., 2012). However, the n-back task emphasizes the updating components involving coding, monitoring, and maintenance (Jaeggi et al., 2010). A comparison of n-back training with WM complex span training showed that n-back training has a more extensive transfer effect in adults (Blacker et al., 2017). In addition, in an intervention study of children with ADHD, compared with the Cogmed program, the running memory task, with a focus on the central executive function component of WM, yielded better effects on WM and ADHD symptoms (Yu et al., 2015). Therefore, the specific training components of different training programs may be the key determining factor to the transfer of WMT to fluid intelligence. Second, previous studies used behavioral indicators to evaluate the transfer effect of WMT in children by comparing the differences between the pre- and post-tests. Nevertheless, behavioral indicators may be insensitive and unable to capture the specific cognitive process and identify the characteristics of information processing in different cognitive stages. Other indices of WMT such as those based on neural responses may be needed to detect the WMT transfer effects, especially when the effects are subtle.
The current study used the event-related potential (ERP) technique to further examine the far and near transfer effects of WMT in preschool children. Previous studies demonstrated that ERP is effective in capturing the temporal dynamics of brain activation patterns in a fine-grained manner, has greater sensitivity to detect differences in executive functioning tasks (Thomas et al., 2022), and is a promising approach to understanding children’s WMC and inhibitory control (Buss et al., 2011, Lotfi et al., 2020). In addition, ERP provides biomarkers for the development of executive function (Downes et al., 2017), and ERP-derived indicators can predict the later development of executive function and academic performance (Harms et al., 2014).
In WMT studies, there are two main categories of ERP signals commonly used to assess the training and transfer effects: exogenously driven components reflecting early attention and perceptual processes, including P1, N1, and P2, and endogenously driven components representing higher-order cognitive processes, such as N2 and P3, which measure inhibitory control, WM, and so on (Wang & Covey, 2020). The P3 amplitude is the positive waveform that appears after 300 ms of the stimulus onset and is intimately related to information processing and context updating (Lenartowicz et al., 2010, Polich, 2007) and can be modified by WMT. An increase in the n-back related P3 amplitude after WMT in healthy adults has been reported (Zhao et al., 2013), and the same increase was observed in 7- to 12-year-old children with dyslexia (Lotfi et al., 2020) and in 10- to 13-year-old children with learning difficulties (Zhang et al., 2018). The P3 amplitude has also been observed in inhibitory control tasks, reflecting inhibitory processes and sustained attention (Grammer et al., 2014). For example, Wang et al., (2022) found that the go/no-go related P3 amplitude was enhanced in children aged 9 to 11 years after WMT.
The N2 component of ERP has also been used to study the transfer effects of WMT on inhibitory control. N2 is a negative electrical waveform following successful inhibition with a peak latency of approximately 200 to 400 ms after stimulus onset. It is generated in the frontal cortex, superior temporal cortex, and anterior cingulate cortex, reflecting conflict-monitoring processes, continuous novel anisotropy, and mismatch recognition (Lo, 2018). For example, recent studies found that after WMT, the N2 amplitude was increased in healthy adults in the n-back task (Covey et al., 2019) as well as in the combined go/no-go flanker task (Wang & Covey, 2020). The N2 effect is also found in inhibitory control tasks, with the high-conflict condition (e.g., incongruent trials in the flanker task, no-go trials in the go/no-go task) generating a greater N2 amplitude than the low-conflict condition (e.g., congruent trials in the flanker task, go trials in the go/no-go task). Although the N2 effect is typically absent in children before 6 years of age (Buss et al., 2011), it has been observed in preschool children after executive attention training (Rueda et al., 2005).
WMT also affects early components of ERP, mainly N1, P1, and P2. These components are exogenously driven and reflect the level of attentional resource investment, the degree of attentional focus on the target stimulus, and so on (Vogel & Luck, 2000). WMT is found to increase N1 amplitude in healthy adults (Wang and Covey, 2020, Zhao et al., 2013).
As mentioned above, because preschool children are in a rapid brain and behavioral developing phase, they are more cognitively and neurologically plastic. WMT in this population is likely to be more efficacious in inducing near and far transfer effects. The current study examined this issue and explored the potential neural mechanisms underlying the WMT effects. Specifically, we explored whether the running memory task, which places a greater emphasis on the component of updating (Pappa et al., 2020), is capable of achieving near and far transfer effects in preschool children. We also tested the idea of whether the updating component in WMT is key to the transfer effects. The running memory task was chosen because it is one of the most common tasks targeting the updating component of WM and is closely related to intelligence (Salthouse, 2014). Peng et al. (2017) demonstrated the transfer of n-back training to fluid intelligence mainly due to its ability to target the updating component. Unfortunately, they did not find the near transfer effect. Reporting only an increase in intelligence scores (far transfer) without identifying an increase in WMC (near transfer) is puzzling given that the effects of WMT on fluid intelligence or academic achievement are thought to be mediated by improvements in WMC (Melby-Lervåg et al., 2016). In the current study, we selected the n-back task and related ERP components as indicators for the near transfer effect, whereas fluid intelligence as assessed in Raven’s Coloured Progressive Matrices was used as indicators for the far transfer effect. Meanwhile, we measured inhibitory control ability as the far transfer effect based on the observations that inhibitory control is closely related to WM and is often used as an important measure of the transfer effect in WMT studies (e.g., Thorell et al., 2009, Wang et al., 2022, Zhang et al., 2019). Thus, we used the go/no-go task and recorded the related ERP components (e.g., St. John et al., 2019). We hypothesized that the running memory training would transfer to preschool children’s WM, inhibitory control, and fluid intelligence. Second, we hypothesized that both the exogenous components of ERP (N1, P1, and P2, reflecting early attention) and endogenous components (P3 and N2, representing WM and inhibitory control) would be enhanced in the post-training tests.
A total of 50 typically developing children from a preschool in Huaian, Jiangsu province, China, participated in this study. They were randomly assigned to the experimental group or the active control group with a balanced grade and gender. There were 26 children in the training group (52–82 months of age, M = 65.65, SD = 8.12; 13 girls) and 24 in the control group (52–75 months of age, M = 63.12, SD = 7.29; 9 girls). None of the children had received a psychiatric diagnosis or participated in