【背景】

工作记忆是指个体对于当下接收到的部分信息进行短暂存储并进行一定程度操作的能力，是人最重要的认知功能之一。工作记忆与人的阅读理解、问题解决等高级认知功能密切相关，因此，越来越多的研究者致力于通过认知训练来改善个体的工作记忆能力。N-back任务和工作记忆广度任务是两种最常用的工作记忆训练范式。既往研究表明，使用视听双维度模式增加n-back训练任务的复杂性可以促进迁移效应的产生。然而，关于双维度工作记忆广度训练是否具有同样的效果目前还不确定。本研究采用随机对照研究设计，开展了为期两个月的双维度工作记忆广度训练和双维度n-back训练，通过行为和神经电生理水平的纵向数据来深入探究工作记忆的训练效应，以便更好的理解训练效果的发生机制，为未来训练任务的开发和训练计划的设计等工作提供必要的理论依据。

【研究方法】

本研究采用随机对照实验设计，共计招募60名健康在校大学生，将其随机分为双维度工作记忆广度任务训练组（8周内完成40次训练）、双维度n-back任务训练组（8周内完成40次训练）和对照组（无干预措施）。所有受训者在训练前（第1次训练前）、训练中（第20次训练结束）、训练后（第40次训练结束）完成三种不同刺激条件的变化觉察任务（4T条件：记住线索提示侧所有四个红色长方形；2T2D条件：记住线索提示侧两个红色长方形，忽略其他两个蓝色长方形；2T2D回溯线索条件：先记住线索提示侧所有红色蓝色长方形，然后根据回溯线索只保留其中的2个红色或蓝色长方形），并记录其行为反应和脑电活动。统计分析时，使用单因素方差分析和卡方检验考察不同组受训者在训练前的匹配性，使用三因素重复测量方差分析考察三组受训者在行为水平（研究一）和电生理（研究二）水平的训练效应，以探索不同训练任务之间作用机制的差异。本研究的显著性水平设定为P<0.05。

【结果】

研究一发现，与对照组相比，接受计算机自适应性双维度工作记忆广度任务训练和双维度n-back任务训练的受训者在具有相同认知成分但范式不同的工作记忆任务（变化觉察任务）上均出现了训练效应。在4T和2T2D回溯条件下，两个训练组的行为表现均有显著提高，表明训练改善了受训者的工作记忆容量；在2T2D条件下，两个训练组的行为表现在训练前后没有显著差异，表明训练未能改善受训者的干扰抑制能力。

研究二发现，与对照组相比，接受双维度工作记忆广度任务训练和双维度n-back任务训练的受训者均出现了脑电指标（CDA）的改善。在4T条件下，CDA结果与行为水平结果一致，表明两种训练均可以增加受训者的工作记忆容量。并且，在4T和2T2D条件，以及2T2D回溯线索条件不同时间窗的比较中，未发现显著的时间x组别x条件的三重交互作用，表明训练未能改善受训者的干扰抑制能力。此外，相比双维度工作记忆广度任务训练，双维度n-back任务训练在2T2D回溯线索条件下CDA的改善更为显著，表明双维度n-back任务对工作记忆容量的改善作用更优。

【结论】

本研究主要探讨了使用双维度工作记忆广度任务和双维度n-back任务的认知训练对工作记忆能力的影响，并对其在行为和电生理两个层面的机制进行了探索。研究结果表明：（1）双维度工作记忆广度任务和双维度n-back任务训练均可以改善个体的工作记忆容量。（2）双维度工作记忆训练策略的使用并不能改善受训者的干扰抑制能力；（3）双维度n-back任务训练比双维度工作记忆广度任务训练对受训者的工作记忆容量具有更显著的改善作用。综合来看，这些发现可以为今后工作记忆训练的应用研究提供更多理论基础，也可以为深入探讨工作记忆的神经基础和训练的可塑性作用提供有意义的启示。

BACKGROUND

Working memory refers to an individual's ability to temporarily store and manipulate a portion of information received in the present moment, and it is one of the most critical cognitive functions in humans. Working memory is closely related to higher-order cognitive functions such as reading comprehension and problem-solving. Therefore, an increasing number of researchers are dedicating efforts to improving individuals' working memory through cognitive training. The N-back task and working memory span task are the two most commonly used working memory training paradigms. Previous studies have shown that increasing the complexity of the n-back training task using audio-visual dual-dimension patterns can promote the transfer effect. However, it is currently unclear whether dual working memory span training has the same effect.

This study employed a randomized controlled trial design to conduct a two-month dual working memory span training and dual n-back training. Through longitudinal data analysis of behavioral and neurophysiological levels, we aim to deeply investigate the effects of working memory training and better understand the underlying mechanisms of the training effect, providing a necessary theoretical basis for the development of future training tasks and the design of training programs.

METHOD

This study employed a randomized controlled trial design and recruited a total of 60 healthy university students. They were randomly assigned to the dual working memory span training group (40 training sessions completed in 8 weeks), the dual n-back training group (40 training sessions completed in 8 weeks), and the control group (without any intervention). All participants completed a change detection task under three different conditions (4T condition: remembering all four red rectangles on the cued side; 2T2D condition: remembering two red rectangles on the cued side and ignoring two blue rectangles; 2T2D retro-cue condition: remembering all the red and blue rectangles on the cued side and then retaining two of them based on a retro-cue) before training (the first training session), during training (after 20 training sessions), and after training (after 40 training sessions), while their behavioral responses and electroencephalogram (EEG) activity were recorded.

For statistical analysis, one-way analysis of variance (ANOVA) and chi-square tests were used to examine the group equivalence before training. Three-way repeated-measures ANOVA was used to examine the training effects of the three groups on both behavioral (study 1) and EEG (study 2) levels, aiming to explore the differences in the mechanisms of the different training tasks. The significance level was set at P < 0.05 in this study.

RESULTS

In Study 1, it was found that compared to the control group, participants who received computerized adaptive dual working memroy span training or dual n-back training showed training effects on a working memory task with the same cognitive components but different paradigms (change detection task). In the 4T and 2T2Dretrocue conditions, both training groups showed significant improvement at behavioral level, indicating that training improved the participants' working memory capacity. In the 2T2D condition, there was no significant difference in behavioral performance before and after training in both training groups, indicating that training did not improve the participants' interference control.

In Study 2, it was found that compared to the control group, participants who received dual working memory span training and dual n-back training showed improvements in brain electrophysiological indices (CDA). In the 4T condition, the CDA results were consistent with the behavioral performance results, indicating that both types of training can increase the participants' working memory capacity. Moreover, in the comparisons of different time windows in the 4T and 2T2D conditions, and the 2T2Dretro condition, no significant triple interaction of time x group x condition was found, indicating that the training did not improve the participants' interference control. Additionally, compared to dual working memory span training, dual n-back training showed a significant improvement in CDA in the 2T2D-backtrack condition, suggesting that dual n-back training had a better effect on improving working memory capacity.

CONCLUSION

The present study found that the gains individuals gained with dual-dimensional training regarding interference with inhibitory abilities did not migrate to a change detection task in a single (visual) dimension. Both dual working memory span task and dual n-back task training can improve an individual's working memory capacity rather than interfere control. In addition, at the neural activity level, the positive effects were greater for training on the dual n-back task than for training on the dual span task.

This study mainly investigated the effects of cognitive training using dual n-back working memory training and dual working memory span training on working memory capacity, and explored the underlying mechanisms at both the behavioral and electrophysiological levels. The results showed that: (1) both dual n-back and dual working memory span training can improve individuals' working memory capacity; (2) the use of dual-task working memory training strategy did not improve the trained participants' ability to interference control; (3) dual n-back training had a more significant effect on improving participants' working memory capacity than dual working memory span training. Taken together, these findings can provide more theoretical foundations for future research on WM training applications, as well as meaningful insights into the neural basis of working memory and the plasticity effects of training.