A prominent but poorly understood domain of human agency is mental action,
i.e., the capacity for reaching specific desirable mental states through an
appropriate monitoring of one's own mental processes. The present paper ai
ms to define mental acts, and to defend their explanatory role against two
objections. One is Gilbert Ryle's contention that postulating mental acts l
eads to an infinite regress. The other is a different although related diff
iculty, here called the access puzzle: How can the mind already know how to
act in order to reach some predefined result? A crucial element in the sol
ution of these puzzles consists in making explicit the contingency between
mental acts and mental operations, parallel to the contingency between phys
ical acts and bodily movements. The paper finally discusses the kind of ref
lexivity at stake in mental acts; it is shown that the capacity to refer to
oneself is not a necessary condition of the successful execution of mental
acts.