这项由 Soha Farboud Sheshdeh 领导、在 Radboud University 的 Donders Institute 进行的研究表明，超声波在眼球运动启动前的极短时段内可微调人的决策。研究采用高速左右选择任务（需在左或右间做快速判断）显示，在决策接近临界点时，一次短促的超声脉冲就能把选择向某一侧推移；其作用在脑区准备启动扫视运动的瞬间发生，延迟为秒的分数量级。该效应在反复试次中稳定出现，尤其在两种选择几乎等势、被试表现出犹豫时最明显，说明外部干预可利用一个非常窄的时间窗影响行为，而非取代大脑的内在决策指令。

研究对象定向于额叶眼场（frontal eye fields），而非全脑照射；结果显示脉冲可在不明显增加广泛扰动的情况下改变该回路的易激性。试验后影像学发现，不同受试者在自然抑制信号的起始水平上差异很大，而该基线差异与超声效应幅度显著相关。该效应与 γ-氨基丁酸（GABA）水平呈持续一致关系：GABA 较高或较低者并非均一受益，说明个体生理差异会放大或减弱调节。另一个核心机制线索来自细胞与机制层面：机械性压力可开闭敏感孔道，扰动胞内钙离子并提高神经元放电概率，但该机制是否普遍适用于所有人仍未定，亦与“头骨会弯折和衰减声波”相关。

在潜在应用上，作者强调这不是“控制他人”而是支持脑功能，但同样提出伦理边界必须清晰。文中也承认，虽然存在一项涉及4次实验的人体研究表明“匹配声学对照”可解释此前被误归因于超声的部分效应，但在这项任务条件下仍观察到稳健结果。未来要进医院、学校或工作场景应用，必须先完成更严格的安全验证，排除加热、微气泡与听觉伪影，建立个体化靶点定位与长期随访框架。该研究已发表于《Nature Communications》，并提出下一步应聚焦标准化靶向、跨被试复制与风险收益监测。

This study led by Soha Farboud Sheshdeh at Radboud University’s Donders Institute reports that a brief transcranial ultrasound pulse can subtly bias a choice in the split second before eye movement. In a fast two-choice task requiring looks to the left or right, the pulse shifted decisions at the tipping point, with effects unfolding over fractions of a second and timing-locked to the pre-saccadic preparation phase when the brain was about to issue the motor command. Across repeated trials the effect was most evident when options were closely matched and participants hesitated, indicating a narrow temporal window in which external input can bias behavior without replacing endogenous command, and suggesting a specific causal influence on decision formation rather than broad arousal or distraction.

The team targeted the frontal eye fields rather than the whole brain, and found the stimulation could alter circuit excitability through the skull at the key moment. In scans taken after choices, baseline inhibitory tone varied strongly across participants, and this baseline predicted how strongly the same pulse changed behavior; the response size tracked GABA levels in a stable participant-wise trend. This implies meaningful person-to-person variability for future interventions. At the cellular level, evidence from neurons suggests mechanical stress may open sensitive membrane pores, alter intracellular calcium, and increase spiking probability, though a single mechanism cannot be assumed across all brains. The authors therefore frame the finding as a precise neuromodulatory effect with a strong timing dependence, not a uniform one-size-fits-all effect.

The paper also stresses ethical constraints: the intervention did not deliver information and only amplified a pre-existing directional bias, so the authors reject the “human robot” framing and emphasize therapeutic support instead. Nonetheless, concerns about misuse or overreach remain, especially if used outside controlled settings. The work also reports a human safety-relevant pattern across four experiments, including findings that matched control sounds explained effects previously attributed to brain-only stimulation; thus auditory confounds must be ruled out. Future progress toward clinics, schools, or workplaces requires strict safety limits (heating, cavitation, acoustic artifacts), standardized targeting, personalized dosing, and long-term monitoring to keep benefit ahead of risk. The study is published in Nature Communications and frames next steps as replication, reproducibility, and scalable, individualized protocols. (Key numbers: 4)