In many task-switch studies, response repetitions produce benefits (response repetition benefit, RR benefit) when the task repeats but produce costs (response repetition costs, RR costs) when the task switches, which was terms as response repetition effects (RR effects). The response repetition is not bound to an exact (i.e., physical) repetition of the same response. Rather, the repetition of an abstract code (i.e. response category) in subsequent trials is sufficient to produce these effects. RR effects can also happen in response modes more than manual response (e.g. verbal response and motor imagery). Even if sequential trials have different response modes also found RR effects. Both stimuli valence and congruency can affect RR effects. The bivalent stimuli, which can be evaluated according to both tasks, yields larger RR costs and smaller RR benefit than univalent stimuli. Compared with incongruent trial, there are greater RR costs and smaller RR benefit after the congruent stimuli. RR costs were larger for bivalent-incongruent stimuli and univalent-incongruent stimuli than for neutral stimuli, which seems to be affected by the congruency of the preceding trial. The more response-related features repeat, the more remarkable the RR effects will be. In addition, longer task preparation time and time pressure would decrease RR benefit, but no consistent effects on RR costs. Three main theoretical frameworks have been proposed to explain these effects. The reconfiguration-based account proposes tasks are organized hierarchically, when a superordinate level (e.g., task-set) switches, this generalized downwards to all subordinate levels (e.g., the response). That is, switching the task implies preparing a different response than the one just executed. This results in costs for repeating the response of the preceding trial, because executing that response required an additional step of reconfiguration to switch back to that response. It provides evidence for neural mechanism, but can seldom explain the mechanisms of different influence factors, which might due to hierarchical control is not a specific process in task switching. The priming and inhibition account supported that responses are inhibited after they have been activated in order to prevent perseveration tendencies. The RR benefits in task-repetition trials are due to the concurrent repetition of the relevant stimulus category. This priming benefit overcompensates the negative effect of response inhibition. Otherwise, no stimulus category priming benefit in task-switch trials when response repeat, which results in RR costs. This theory was always studied in dual-task, which requires to distinguish from the study in task switching. According to the episodic-retrieval account, all task features (e.g. cue, task set, stimulus and response) serve as retrieval cues for previous episodes in which these features have been encountered and bound. Whenever at least one feature of a previous episode repeats, the previous event file is retrieved. Such retrieval is helpful while all features of the previous episode repeat in current episode. However, if only some features of the previous episode repeat, performance decreases considerably, due to the bindings mismatching and new bindings have to be established. The account only focuses on exact repetition but no abstract code repetition, or it is not clear which features an episode consists of. Furthermore, the weighting of different retrieval cues and matching and mismatching features need several additional assumptions. Current neuroscience study only focused on the neural basis and theoretical framework, more elaborate and concreted mechanisms should be discussed. To further explore various sources for RR effects is required and should base on integrating different theories.