Growing evidence for moment-to-moment fluctuations in visual attention has led to questions about the impetus and time course of cognitive control. These questions are typically investigated with paradigms like the flanker task, which require participants to inhibit an automatic response before making a decision. Connectionist modeling work suggests that between-trial changes in attention result from fluctuations in conflict-as conflict occurs, attention needs to be upregulated to resolve it. Current sequential sampling models (SSMs) of within-trial effects, however, suggest that attention focuses on a goal-relevant target as a function of time. We propose that within-trial changes in cognitive control and attention are emergent properties of the dynamics of the decision itself. We tested our hypothesis by developing a set of SSMs, each making alternative assumptions about attention modulation and evidence accumulation mechanisms. Combining the SSM framework with likelihood-free Bayesian approximation methods allowed us to conduct quantified comparisons between subject-level fits. Models included either time- or control-based attention mechanisms, and either strongly- (via feedforward inhibition) or weakly correlated (via leak and lateral inhibition) evidence accumulation mechanisms. We fit all models to behavioral data collected in variants of the flanker task, one accompanied by EEG measures. Across three experiments, we found converging evidence that control-based attention processes in combination with evidence accumulation mechanisms governed by leak and lateral inhibition provided the best fits to behavioral data, and uniquely mapped onto observed decision-related signals in the brain. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

中文翻译：

---

用于决策的动态审判内冲突解决模型。

越来越多的证据表明视觉注意力的瞬间波动导致了关于认知控制的动力和时间过程的问题。这些问题通常使用像侧翼任务这样的范式进行调查，这些范式要求参与者在做出决定之前抑制自动响应。联结主义建模工作表明，试验间注意力的变化是由冲突的波动引起的——当冲突发生时，注意力需要上调来解决它。然而，当前试验内效应的顺序抽样模型 (SSM) 表明，注意力集中在与目标相关的目标上，作为时间的函数。我们提出，认知控制和注意力的试验内变化是决策本身动态的紧急特性。我们通过开发一组 SSM 来检验我们的假设，每个人都对注意力调制和证据积累机制做出了替代假设。将 SSM 框架与无似然贝叶斯近似方法相结合，我们可以在主题级别的拟合之间进行量化比较。模型包括基于时间或控制的注意力机制，以及强（通过前馈抑制）或弱相关（通过泄漏和侧向抑制）证据积累机制。我们将所有模型拟合到在侧卫任务的变体中收集的行为数据，其中一个伴随着 EEG 测量。在三个实验中，我们发现了汇聚证据，即基于控制的注意力过程与由泄漏和侧向抑制控制的证据积累机制相结合，提供了对行为数据的最佳拟合，并唯一地映射到大脑中观察到的与决策相关的信号上。（PsycInfo 数据库记录 (c) 2020 APA，保留所有权利）。