【研究背景】
工作记忆是对目标信息进行暂时性储存的容量有限的系统，是人类最基本的认知功能之一。工作记忆在很大程度上制约了阅读、计算、推理等其他更高级的认知功能。其缺陷被认为是很多精神疾病如精神分裂症、注意缺陷多动障碍和阿尔兹海默症等的核心认知缺陷之一，不仅在一定程度上影响了患者的临床症状，还是疾病预后和转归的最强有力预测因素之一(Li,2016; Gropper,2014;)。在当前缺乏可有效改善工作记忆的化学药物的现状下，开发物理训练可能是最快捷和有效途径之一。其中，电子游戏类工作记忆训练备受关注。
工作记忆广度任务（对顺序随机出现的若干刺激进行记忆）和变化觉察任务（对同时出现的若干刺激进行记忆）是常用的工作记忆容量评测范式。事实上，工作记忆广度任务已经作为工作记忆训练范式获得了广泛应用。但遗憾地是，正如一些meta分析所指出的，目前并不确定工作记忆广度任务是否可以有效改善工作记忆过程，原因是既往研究未使用非工作记忆广度任务作为评价任务。尤其，从治疗疾病的工作记忆缺陷出发，CNTRICS项目已经提供了工作记忆缺陷的评估工具，但既往研究并未探讨工作记忆广度任务是否有效改善这些评估工具的得分。
本研究主要从这一实际问题出发，对单一工作记忆广度任务训练对变化觉察任务和其他CNTRICS项目推荐任务的行为和/或电生理效果、效果的持久性及其间个体差异性的影响因素进行了系列研究。
【研究方法】
研究一采用了随机对照的实验设计，一共招募了80例健康大学生被试，将其随机分为自适应性训练组（所有被试在4 周内完成20次视空间工作记忆广度任务训练）和控制组（所有被试在4 周内完成20次低难度视空间工作记忆广度任务的重复性练习）。在训练前后，被试分别完成脑电任务和行为任务。脑电任务为DPX任务和低难度变化觉察任务（记忆2个彩色方块）。行为任务为空间广度任务的测试版和变化觉察任务的行为版。研究中使用重复测量方差分析的方法，比较两组被试训练前后在行为和电生理两个水平上的差异。在分析中，时间为被试内因素，组别为被试间因素。
研究二同样采用了随机对照的实验设计方法，一共招募了60例健康大学生，将其随机分为自适应性训练组（干预方案同研究一）和控制组（干预方案同研究一）。与实验一不同的是我们在研究二中增加了任务难度，同时也对训练效果进行了追踪。在训练前、训练后、训练后2个月（随访1）和训练后6个月（随访2）时采集所有被试在完成高难度变化觉察任务（记忆4个彩色方块）时的行为和电生理数据。采用了重复测量方差分析来比较两组被试在不同的时间点、行为水平和电生理水平上的差异。在分析中，时间为被试内因素，组别为被试间因素。
研究三首先对研究一（N = 40）和研究二（N = 30）训练组的电生理数据进行分析。根据训练前变化觉察任务的成绩估计被试的工作记忆容量，然后根据工作记忆容量的均值，把被试分为高容量和低容量两个组。采用重复测量的方差分析（时间为被试内因素，组别为试间因素，即根据被试的工作记忆容量来进行区分组别，划分标准为工作记忆容量的中位数）比较两组被试在训练前、训练后和随访时的电生理差异。在此基础上，招募大量健康大学生（N = 460）进行自适应性训练（所有被试在4 周内完成20次视空间工作记忆广度任务训练），并在训练前、训练后和训练后6个月时接受多任务工作记忆行为水平测评。使用重复测量的方差分析（被试内因素是时间，被试间因素是组别）比较两组被试训练前、训练后和随访时的行为差异。
【研究结果】
研究一为空间工作记忆广度任务训练改善工作记忆目标维持能力提供了电生理证据。通过重复测量方差分析对比训练组与控制组在训练前和训练后两个时间点上的差异，未能在行为水平发现显著的时间 ? 组别交互作用，但在电生理水平获得阳性结果，主要包括在DPX任务延迟阶段诱发的CNV（F (1,67) = 4.765，P = 0.033）和低难度变化觉察任务延迟阶段诱发的CDA（F (1,68) = 4.286，P = 0.042）上发现显著的时间 ? 组别交互作用。进一步事后配对t检验显示训练组的CNV（t(34) = 2.923，P = 0.006）和CDA（t (33)= 2.152，P = 0.039）显著增加，控制组无显著性改变（Ps > 0.05）。
研究二为空间工作记忆广度任务训练改善工作记忆容量提供了行为和电生理证据，但该作用并不持久。通过重复测量方差对比训练组与控制组在训练前和训练后两个时间点上的差异，在工作记忆容量估计值K分（F (1,53) = 5.352，P = 0.025）和高难度变化觉察任务延迟阶段诱发的CDA（F(1,42) = 4.644 , P = 0.037）上均发现显著的时间 ? 组别交互作用。进一步事后配对t检验显示训练组的K分（F(1,53) = 26.81，P < 0.001）和CDA（F(1, 21) = 4.731，P = 0.041）显著增加，控制组无显著性改变（Ps > 0.05）。但，当考虑随访时的行为和电生理指标时，未发现任何显著的时间 ? 组别交互作用（Ps > 0.05）。
研究三为训练前认知水平对训练效果和训练效果持久性的影响提供了行为和电生理证据。首先，对研究一训练组在训练前和训练后两个时间点完成低难度变化觉察任务的电生理数据进行重复测量的方差分析，结果发现了显著的时间 × 组别的交互作用（F (1,32)= 5.209，P = 0.030）。进一步事后配对t检验发现低容量组训练后CDA显著增大（t (1,16) = 8.771，P = 0.010），而高容量组无显著改变（t (1,16)= 0.084，P = 0.776）。然后，对研究二训练组在训练前、训练后、训练后6个月三个时间点上完成高难度变化觉察任务的电生理数据进行重复测量的方差分析，我们同样发现了显著的时间 × 组别的交互作用（F(1,20) = 4.584，P = 0.025），进一步事后配对t检验发现，低容量组训练后CDA显著增大（t (1,10) = 4.605, P = 0.001），在6个月随访时降低至与训练前相似水平（t(1,10)  = 1.530, P = 0.157），高容量组无显著性改变（Ps > 0.05）。最后，通过对432例健康大学生在训练前、训练后和训练后6个月的行为数据进行重复测量的方差分析，发现训练前工作记忆容量水平会影响训练的效果和持久性。对K分的分析发现显著的时间 ? 组别的交互作用（F(1,430) = 31.455，P < 0.001）。进一步事后配对t检验发现，低容量组Kmax在训练后得到显著增加（t(1,213) = -12.421，P < 0.001），且这一效应持续至训练结束后6个月（t (1,213)= -9.430，P < 0.001），而高容量组Kmax在训练后也获得显著增加（t (1,217)= -4.910，P < 0.001），但这一效应不能持续至训练结束后6个月（t (1,217)= -1.204，P = 0.230）。但是对其他任务的分析结果显示，没有显著的时间 × 组别的交互作用（Ps > 0.05）。
【研究结论】
本研究通过开展一系列行为和电生理水平的研究证明了工作记忆广度任务训练可以作为提升工作记忆目标维持能力（研究一）和工作记忆容量（研究二）的有效工具，但训练效果并不持久（研究二）。训练前认知水平是影响训练效果和训练效果持久性的重要因素（研究三）。这些结论将为未来完善工作记忆广度训练方案提供理论依据。

【BACKGROUND】
Working memory is a system with limited capacity for temporary storage of target information, which is one of the most basic cognitive functions of human beings. It restricts reading, calculation, reasoning and other higher cognitive functions to a great extent. Its defect is considered to be one of the core cognitive defects of many mental diseases, such as schizophrenia, attention deficit hyperactivity disorder and Alzheimer's disease. It not only affects the clinical symptoms of patients to a certain extent, but also is one of the most powerful predictors of disease prognosis and outcome. In the current situation of the lack of chemical drugs which can effectively improve the working memory, the development of physical training may be one of the fastest and most effective ways. Among them, working memory training of video games has attracted much attention.
Working memory span task (to remember several stimuli that appear randomly in sequence) and change awareness task (to remember several stimuli that appear at the same time) are commonly used working memory capacity assessment paradigms. In fact, working memory span task has been widely used as working memory training paradigm. Unfortunately, as pointed out by some meta-analyses, it is uncertain whether working memory span task can effectively improve working memory process, because previous studies did not involve non-span tasks in their evaluation. In particular, from the perspective of the treatment of working memory deficits in mental disorders, although CNTRICS project has provided assessment tools for their working memory deficits, previous studies have not explored whether the working memory span task training can effectively improve the scores of these assessment tools.
Mainly due to this practical problem, this study using behavioral and / or electrophysiological assessment during change detection task and/or other CNTRICS tasks to test the effects of working memory span task training (study 1 and study 2), the persistence of the effects (study 2) and the influencing factors of individual differences for the effects and effect persistence (study 3).
【METHODS】
The tudy 1 was a randomized controlled trial. A total of 80 healthy college students were recruited and randomly divided into adaptive training group (all subjects completed visuospatial working memory span task training) and non-adaptive control group (all subjects repeatedly practice on low difficult visuospatial working memory span task). both before and after the training, electrophysiological data of all subjects were collected when they completed the DPX task and an easy version of the change detection task (remember 2 color squares). Repeated measurement ANOVA was used to compare the differences of behavior and electrophysiology between the two groups before and after training. In this analysis, the intra subject factor is time, and the inter subject factor is group.
The study 2 was also designed as a randomized controlled trial. A total of 60 healthy college students were recruited. They were randomly divided into adaptive training group (the same intervention scheme as that in study 1) and non-adaptive control group (the same intervention scheme as that in study 1). Behavioral and/or electrophysiological data were collected at pre-test, post-test and follow-up (2 months after training, follow-up 1; 6 months after training, follow-up 2). Repeated measurement ANOVA was used to compare the differences of behavior and electrophysiology between the two groups across the training. In this analysis, the intra subject factor is time, and the inter subject factor is group.
In study 3, the electrophysiological data of the training group within both study 1 (n = 40） and study 2 (n = 30) were analyzed first. Subjects' working memory capacity was estimated  according to the performance (Kmax) of change detection task before the training. Subjects were then divided into high capacity group and low capacity group according to the median of Kmax. The electrophysiological differences between the two groups at pre-test, post-test and follow-up (6 months after training) were compared using repeated measures ANOVA (internal factor of subjects was time, and inter subject factor was group). Afterwards, a large number of healthy college students (n = 432) were recruited to carry out adaptive training (all subjects completed visuospatial working memory span task training). Multiple working memory tasks were evaluated at pre-test, post-test and follow-up (6 months after the training). The behavioral differences between the two groups across the three time points were compared using repeated measurement ANOVA (the internal factor of the subjects was time, and the inter subject factor was group).
【RESULTS】
The study 1 provided electrophysiological evidence for the improvement of working memory span task training on working memory goal maintenance procss. No significant time ? group interaction was found at the behavioral level, but positive results were obtained at the electrophysiological level, which included the CNV (F (1,67) = 4.765, P = 0.033) that was induced by DPX task during the delay period and the CDA (F(1,68) = 4.286, P = 0. 042) that was induced by the easy version of the change detection task. Further paired t-test showed that both CNV (t (34) = 2.923, P = 0.006) and CDA (t (33)= 2.152, P = 0.039) were significantly increased in the training group, while there was no significant change in the control group (PS > 0.05).
The study 2 provided both behavioral and electrophysiological evidence for the improvement of working memory span task training on working memory capacity, but the effect was not lasting. The repeated measurement ANOVA on the difference between the training group and the control group at pretest and posttest found significant time ? group interaction at both K score (F(1,53) = 5.352, P = 0.025) and CDA (F (1,42) = 4.644, P = 0.037). Further post-hoc paired t-test showed that K score (F  (1,53) = 26.81, P < 0.001) and CDA (F(1, 21)  = 4.731, P = 0.041) in the training group were significantly increased, while no significant changes were found in the control group (PS > 0.05). When the behavioral and electrophysiological parameters at follow up were considered, no significant time x group interaction was found (PS > 0.05).
The study 3 provides both behavioral and electrophysiological evidence for the influence of pretest cognitive level on training effect and training effect persistence. The repeated measurement ANOVA was used to analyze the electrophysiological data of the easy version of the change detection task between the subjects with high capacity and the subjects with low capacity in Study 1 at pretest and posttest. Significant time × group interaction was found (F(1,32) = 5.209, P = 0.030). Further paired t-test showed that CDA increased significantly in the low capacity group after training (t (1,16) = 8.771, P = 0.010), but not in the high capacity group (t (1,16)= 0.084, P = 0.776). The repeated measurement ANOVA was also used to analyze the electrophysiological data of the difficult version of the change detection task between the subjects with high capacity and the subjects with low capacity in Study 2 at pretest, posttest and follow-up (6 months after the training). Significant time × group interaction (F (1,20) = 4.584, P = 0.025) was also found. Further, paired t-test showed that CDA in low capacity group increased significantly after the training (t (1,10)= 4.605, P = 0.025), however, it decreased at the follow-up to similar level as pretest (t(1,10) = 1.530, P = 0.157). By contrast, high capacity group did not show any significant changes (PS > 0.05). Finally, the repeated measurement of ANOVA within 432 healthy college students' behavior data at pretest, posttest and follow-up (6 months after training) found that working memory capacity before the training affects the training effect and persistence. Significant time ? group interaction was found at Kmax (F(1,430) = 31.455, P < 0.001). Further paired t-test showed that Kmax in low capacity group was significantly increased after the training (t (1,213)= -12.421, P < 0.001), and this effect lasted till the follow-up (t (1,213)= -9.430, P < 0.001). By contrast, Kmax in high capacity group also increased significantly after the training (t (1,217)= -4.910, P < 0.001), but this effect could not last till the follow-up (t(1,217) = -1.204, P = 0.230). No significant time × group interaction was found in other tasks (PS > 0.05).
【CONCLUSIONS】
Through a series of behavioral and electrophysiological substudies, this study suggests that working memory span task training can be used as an effective tool to improve working memory goal maintenance ability (Study 1） and working memory capacity (Study 2）, but the training effect is not lasting (Study 2）. The cognitive level before the training is an important factor affecting the training effect and the persistence of training effect (Study 3. These conclusions will provide a theoretical basis for improving the working memory span training program in the future.