压力伴随着每个人生活工作和毕生发展。适度应激（俗称压力）能够激发我们机体防御的机制，促进环境适应与生存能力。然而，超负荷压力是引发全球焦虑与抑郁等精神健康问题发病率持续恶化的关键风险因素。当面临长期压力时，有的人总能克服重重困难并不断提升自我调节与应对能力，有的人却表现不佳、发挥失常，甚至引发心理问题或障碍。那么，究竟是什么因素或特质影响个体长期压力下以上不同表现呢？其中，特质焦虑，即对不确定性环境或者潜在威胁刺激往往表现出敏感性焦虑反应倾向，被认为是一种关键的调控因素之一。然而，其背后的认知神经机理不明：当高焦虑特质和低焦虑特质人群同时面临长期外部压力时，他们究竟会有哪些不同的认知行为表现及其神经活动模式？长期压力究竟是苦剂还是良药？

在本研究中，我们招募了36名为即将到来的竞争激烈的中国研究生入学考试准备了至少6个月的被试进入到长期压力组。在对照组中，招募了34名在过去6个月内没有为中国研究生入学考试做准备并且没有暴露于其他主要压力源的年龄和教育相匹配的被试。在两组中分别测量每个被试的特质焦虑与心理压力水平。我们使用了经典的n-back工作记忆任务和功能性磁共振成像，利用贝叶斯层级漂移扩散模型对行为数据建模，同时在激活和网络水平上分析脑像数据，探讨了长期压力和特质焦虑如何通过改变脑网络平衡来影响n-back任务执行过程中的动态认知计算过程。

研究发现，与对照组相比，长期压力下的被试体验到更高的心理压力水平，在执行n-back任务时表现出更快的证据积累速度和更低的决策阈值，但其准确性与控制组相当。当考虑到个体的特质焦虑后，较低的特质焦虑个体在长期压力下表现出相对更快的漂移速度，而准确性在不同群体和焦虑水平上保持在相当水平。在脑像水平上，全脑2 (压力组、控制组) x 2 (0-back、2-back)方差分析的组别主效应进一步揭示了在长期压力下，前脑岛的激活增加。组别和工作记忆负荷的交互效应显示与对照组相比，当任务负荷增加时，压力组的默认模式网络的负激活显著减少，而突显网络过度负激活。多组结构方程模型揭示了额顶活动在特质焦虑和漂移率的间接关系中起到显著的中介作用 —— 在长期压力组而非控制组中，具有较高特质焦虑的个体在具有更高认知要求的任务条件下，通过更高的额顶叶网络的活动降低了漂移率。最后，与高任务需求的对照组相比，长期压力还导致额顶网络-默认模式网络间的功能连接解耦更强，并且更高的特质焦虑预示着更强的额顶网络-默认模式网络间的功能连接解耦。

综上，我们的研究结果提出了一个神经认知模型，揭示了长期压力和特质焦虑如何通过相互作用来改变适应性和适应不良的执行认知功能从而影响工作记忆过程中的潜在动态决策过程，为心理应激与情感认知相互影响领域提供了新的见解：即长期压力能激发低特质焦虑人群的潜力从而更好的完成任务；而高特质焦虑本身并不一定导致认知缺陷。相反，高特质焦虑与长期压力协同作用，通过募集更多的认知资源以降低速度为代价来完成当前任务，从而进一步为与压力相关的适应性行为提供个性化评估和预防措施。

Stress accompanies everyone's life, work and development. Moderate stress can stimulate our body's defense mechanisms, and further promote its environmental adaptation and survival. However, excessive stress is a key risk factor for worsening global incidence of mental health problems such as anxiety and depression. Under long-term stress, some people would always overcome numerous difficulties and improve their self-regulation and coping ability, while others exhibit the exact opposite way, and even develop psychological problems. Which factor or trait biases individual’s performance under long-term stress? Trait anxiety, a stable disposition to interpret a wide range of environmental events in a negative way, is recognized as a vulnerable factor accounting for stress-induced adaptive or maladaptive effects. However, the underlying cognitive neural mechanism is unclear: when people with high and low anxiety traits face long-term external stress at the same time, how will their cognitive behavioral and neural activity patterns diverge? Is chronic stress a bitter pill or a cure?

In this study, 36 participants of long-term stress group were recruited from those who have been preparing for the upcoming competitive Chinese National Postgraduate Entrance Exam (CNPEE) for at least 6 months. In the control group, 34 participants matched in age and education who were not preparing for the CNPEE and did not have exposure to other major stressors in past 6 months were recruited. Trait anxiety and psychological distress were assessed one day prior to the fMRI experiment. A Bayesian hierarchical version of the DDM (HDDM) was implemented to estimate latent dynamic decision parameters during task processing. Brain activation and network approaches were employed to identify how long-term stress and trait anxiety alter functional brain network balance.

Our study found that in comparison to controls, male participants under long-term stress experienced higher psychological distress and exhibited faster evidence accumulation but had a lower decision-threshold when performing n-back tasks in general. When taking individual’s trait anxiety into account, lower trait-anxious individuals exhibited relatively faster drift rate in those under long-term stress than in controls, while accuracy remained at a comparable level across groups and anxiety levels. Whole-brain 2 (Group)-by-2 (WM-load) ANOVA revealed a hyper-activation in the anterior insula from the main effect of Group. We also observed a Group-by-WM interaction in regions of the SN and DMN, with greater WM-related activation in the SN and less WM-related deactivation in the DMN in the long-term stress group than controls. Multi-group structural equation models revealed significant mediation effects of the frontoparietal activity on the indirect association between trait anxiety and drift rate under long-term stress but not controls. The results indicate that individuals with higher trait anxiety are prone to exhibit slower evidence accumulation than controls, mediated through higher frontal-parietal activity during high task demand under long-term stress. Long-term stress also led to stronger DMN decoupling with the FPN than controls in high task demand, with higher trait anxiety predictive of stronger FPN-DMN decoupling in those under long-term stress.

To sum up, our findings suggest a neurocognitive model of how long-term stress and trait anxiety interplay to affect latent dynamic computations in executive functioning with adaptive and maladaptive changes. We provide new insight into the interaction between psychological stress and affective cognition: long-term stress can trigger the potentials of low-trait anxious individuals to perform tasks better, and high trait anxiety itself does not necessarily lead to cognitive deficits. Instead, high-trait anxiety works in tandem with chronic stress to complete the task at the expense of slower performance by recruiting more cognitive resources. Our results further inform personalized assessments and preventions for stress vulnerability.