The absence of a necessary connection in singular causation is a key step in the Humean argument against any form of necessity in causation. I argue that Hume's defence of this step is unsuccessful, and that the step could be skipped, accepting the possibility of necessary a posteriori truths. Still this does not suffice to guarantee a necessary connection in singular causation. Necessary a posteriori truths should be backed by necessary a priori truths. Thus, a main object of this paper is to argue that an a priori philosophical concept of causality involves a necessary connection between its terms.