Abstract: Task switching is an important component of executive functions. Switch costs refer to the differences in reaction times and error rates between switch and repeat conditions, theoretical explanations of which include task-set inertia, task-set reconfiguration, and associative retrieval. The task-set inertia account claims that task-set still remains after a task has been implemented, and this task-set inertia is the key element to give rise to switch costs. Strong evidence for the task-set inertia account came from studies that found n-2 repetition costs (backward inhibition), i.e. longer reaction times in n-2 repetition trials (ABA) than in n-2 switch trials (CBA). In recent years, the task-set inertia account has been supported by new evidence, but the nature of inhibition was questioned. Decrease of alpha power during task switching confirmed that inhibition is an important process of task switching. Nevertheless, some studies replicated n-2 repetition costs, but pointed out the nature of backward inhibition might be proactive control, not reactive control as traditional views deemed. The task-set reconfiguration account proposes that endogenous control is needed to change tasks, which may include shifting attention between task-sets, retrieving task goals and rules into working memory, enabling a different response-set. Main evidence for the reconfiguration account came from studies that found preparation effect, an effect that switch costs decrease as preparation time increases. However, switch costs cannot be fully eliminated by preparation, and thus the cause of the residual switch costs must be probed. There are two reconfiguration models to explain the residual switch costs: two-stage model and all-or-none model. In recent years, the reconfiguration account has been supported by a series of studies. Studies that explored hierarchical effect in task switching indicated reconfiguration is a necessary process as it takes longer time to switch upward than switch downward. Some studies compared switch costs following preparation and following performance, and found that preparation alone is sufficient to drive subsequent switch costs, but performance drives more, consistent with the two-stage model. The reconfiguration account has also been supported by the increased theta power in switch conditions, since theta power is supposed to reflect top-down cognitive control. According to the associative retrieval account, switch costs are the results of competition between potentially relevant tasks, e.g. competition between different S-R associations. In recent years, several studies have developed the associative retrieval account, revealing sizable effects of stimulus repetition, response repetition and cue repetition, and controlling of these effects reduced switch costs. In other words, episodic retrieval and associative learning may play an important role in task switching. All of these views have been supported by new studies in recent years, yet none is sufficient to explain all the effects of task switching. Although the task-set reconfiguration account and the task-set inertia account diverge on the issue of proactive control and reactive control, both of them acknowledge the role of higher-order control processes during task switching, while the associative retrieval account challenges it and favors a simpler explanation. To resolve the controversies, the task-set reconfiguration account and the task-set inertia account should be examined more strictly, with episodic retrieval confounds controlled, and the associative retrieval account needs more experimental evidence, especially evidence for neural mechanisms. It is most likely that all these views tell partial understanding of a full story. Therefore, integrated models that reflect influences from key parameters may be constructed in the future.

Key words: task switching, switch costs, task-set inertia, task-set reconfiguration, associative retrieval

摘要： 任务切换是研究认知控制的主要范式之一。大量研究发现切换试次比重复试次的反应时更长，错误率更高，这种差异称为切换代价。任务切换时所产生切换代价的理论解释主要有惯性论、重构论和联结论。近十年来，这些理论均获得新的实验支持和发展，但其争议依旧，没有哪一理论能成功解释任务切换的所有效应。未来研究可以建立整合模型，以准确描述切换代价产生的认知机制。

关键词: 任务切换, 切换代价, 任务设置惯性, 任务设置重构, 刺激-反应联结竞争