If my hand lays a club forcefully on the nearest bystander, we can indeed say that this motion resulted from an inner club-wielding desire of mine; but we must add that I had nothing to do with that desire, and that it arose, to be followed by its inevitable effect, no less to my astonishment than to his. Things like this do, alas, sometimes happen. We are all sometimes seized by compulsive impulses that arise we know not whither, and we do sometimes act upon these. But since they are far from being examples of free, voluntary and responsible behavior, we need only to learn that behavior was of this sort to conclude that it was not free, voluntary, nor responsible. It was erratic, impulsive, and irresponsible. (p. 48)
As long as this criticism is limited to the idea that our desires might arise completely at random or without cause, this conclusion seems plausible. However, when Taylor suggests this sometimes actually occurs, one would like to see more argument in its defense. And in this case Taylor draws an unwarranted conclusion based on two meanings of "responsible".
Taylor means to imply that behavior is not free when it arises from an unknown source within ourselves. In any normal sense such behavior would be voluntary--since it is based on desires--and it is apparently also free and behavior for which one might be morally responsible. So, how can this be behavior for which one would not be responsible? Only because this behavior is irresponsible. Here's the trick; the argument relies on an equivocation on the meaning of "responsible". One sense "responsible" means behavior for which one can reasonably praised or blamed. The other sense of "responsible" is a normative claim, a claim that one is acting in a morally good way or in a way that one should. So, when Taylor's premise is that this behavior is irresponsible, he can only mean it in the second sense since such actions really would be morally unacceptable. They would lack the kind of reflection and care we would consider necessary for morally good action. But his conclusion is about the first sense since that sense coincides with voluntary and free behavior. So Taylor trades on the two meanings of "responsible" in order to support the conclusion that one cannot be morally responsible for the action--in the sense that one can reasonably be praised or blamed for it--because the action is irresponsible--in the sense that such an action is morally bad. If we keep track of the difference in meanings, we can see that the conclusion does not follow.
This does raise the question of whether, and if so why, we can be morally responsible for actions that do not arise from our settled character but which arise spontaneously in some way. But that's an issue that requires more thought.