Recent task-switching studies accumulated evidence in favour of feature binding processes. One marker thereof is the response-speed benefit in full repetition trials, where task, response, and a task-irrelevant feature (i.e., context) repeat, compared to partial repetitions where the context switches. This is explained by reoccurring features retrieving the binding of the previous trial. Previous studies indicate that the benefit is higher when the context appears early (before the target) versus late in a trial (with target onset). However, in most of these studies, an early context was also presented for a longer time. In the present study, we manipulated the context onset (before, with, and after cue onset) while keeping context presentation duration constant. In Experiment 1 (N=95), we did not observe full repetition benefits, nor a modulating effect of context onset thereof. In Experiment 2 (N=96), we observed full repetition benefits, and these were modulated by context onset. Specifically, a repeating context seems to retrieve the previous-trial task and response only when it appeared simultaneous to the cue. This suggests that context onset plays a role in episodic retrieval, independently from context presentation duration. Furthermore, we observed a contextual modulation of the response repetition cost which we interpreted in terms of a combination of binding and response inhibition. Taken together, this study further specifies the dynamics of binding and retrieval of task irrelevant contextual features in task switching.