It is one thing to say that an agent S has a duty to perform action x, and it is quite another thing to say that S fulfills or discharges a duty by performing x. If S has a duty to perform x, of course, then S's performance of x fulfills this particular duty, other things being equal. But if S fulfills a duty by performing x, then it does not follow that S has (or had) a duty to perform x. For it might be that it is by virtue of having a different duty, such as a duty to perform either action x or action y, that S fulfills a duty by performing x. If I have a duty to rescue one of two drowning persons, it not being in my power to rescue both of them, then I fulfill this duty by rescuing one of them. But I do not have a duty to rescue the particular one I rescue, for I cannot reasonably be charged with a failure to discharge my duty in the event that I choose to rescue the other.
Although this distinction has been long recognized by moral philosophers, it has proven to be surprisingly easy to run afoul of in the context of discussing supererogation. An act of supererogation is standardly described as an act (i) Which the agent performing it has no duty or obligation to perform (or, equivalently, which is not obligatory for the agent to perform), (ii) Whose performance by the agent is praiseworthy or meritorious, and (iii) Whose omission by the agent is neither blameworthy nor forbidden. 1 The first of these conditions I shall refer to as the "duty condition", and I will argue that the duty condition, as standardly described, 2 leads to serious difficulties. I shall then suggest that a reformulation of the duty condition in the language of fulfilling duties is a more promising alternative.
In an insightful article which has received regrettably little attention, Michael Stocker begins with the following observation: