We infer rather than perceive the moment we decided to act.

Banks and Isham do a further followup (see this post for a previous followup) on the famous Libet experiment that showed that the reported time of a decision to perform a simple action is at least 300 ms after the onset of brain activity that normally precedes the action. They propose that the reported time is not uniquely determined by any generator of the readiness potential (the commonly head view), but rather is the time participants select on the basis of available cues, chief among them being the apparent time of response. From their abstract:

In Experiment 1, we presented deceptive feedback (an auditory beep) 5 to 60 ms after the action to signify a movement time later than the actual movement. The reported time of decision moved forward in time linearly with the delay in feedback, and came after the muscular initiation of the response at all but the 5-ms delay. In Experiment 2, participants viewed their hand with and without a 120-ms video delay, and gave a time of decision 44 ms later with than without the delay. We conclude that participants' report of their decision time is largely inferred from the apparent time of response. The perception of a hypothetical brain event prior to the response could have, at most, a small influence.