Interactions between rhythmic and feature predictions to create parallel time-content associations

The brain is inherently proactive, constantly predicting the when (moment) and what (content) of future input in order to optimize information processing. Previous research on such predictions has mainly studied the ‘when’ or ‘what’ domain separately, missing to investigate the potential integration of both types of predictive information. In the absence of such integration, temporal cues are assumed to enhance any upcoming content at the predicted moment in time (general temporal predictor). However, if the when and what prediction domain were integrated, a much more flexible neural mechanism may be proposed in which temporal-feature interactions would allow for the creation of multiple concurrent time-content predictions (parallel time-content predictor). Here, we used a temporal association paradigm in two experiments in which sound identity was systematically paired with a specific time delay after the offset of a rhythmic visual input stream. In Experiment 1, we revealed that participants associated the time delay of presentation with the identity of the sound. In Experiment 2, we unexpectedly found that the strength of this temporal association was negatively related to the EEG steady-state evoked responses (SSVEP) in preceding trials, showing that after high neuronal responses participants responded inconsistent with the time-content associations, similar to adaptation mechanisms. In this experiment, time-content associations were only present for low SSVEP responses in previous trials. These results tentatively show that it is possible to represent multiple time-content paired predictions in parallel, however future 32 research is needed to investigate this interaction further.

大脑天生具有前瞻性，会持续预测未来输入信息的时机（when）与内容（what），以此优化信息处理流程。过往针对此类预测的研究大多仅单独探索“时机”或“内容”单一维度的预测机制，未能同时考察这两类预测信息的潜在整合方式。若未实现此类整合，时序线索仅会在预测时刻增强所有即将到来的内容（通用时序预测器（general temporal predictor））。但若将“时机”与“内容”的预测维度加以整合，则可提出更为灵活的神经机制：其时序-特征交互作用能够支持生成多组并行的时序-内容配对预测（并行时序-内容预测器（parallel time-content predictor））。本研究通过两项实验采用时序关联范式：在节律性视觉输入流偏移后，将声音身份与特定时长的延迟进行系统性配对。实验1结果显示，被试会将刺激呈现的延迟时长与声音身份建立关联。实验2中我们意外发现，此类时序关联的强度与前一试次中的脑电图稳态诱发电位（SSVEP）强度呈负相关：这表明当神经元响应较强时，被试的行为表现与时序-内容关联不一致，该现象与适应机制类似。在本实验中，仅当前一试次的SSVEP响应强度较低时，才会存在时序-内容关联。本研究结果初步表明，并行表征多组时序-内容配对预测具备可行性，然而未来仍需开展进一步研究以深入探索此类交互作用。