Utilitarian Generalization (UG) is the theory that an act is right if and only if the consequences of everyone's performing that sort of action would be at least as good as the consequences of everyone's performing any alternative sort of action. A standard issue raised by the theory is how we specify what 'sort' an action is for purposes of applying the theory. According to David Lyons, the action's 'sort' is determined by all and only its 'consequentially significant' properties, i.e., those in virtue of which it produces utilities or disutilitiesJ This implies that in certain cases the action's sort must refer to the activities of other agents, for they form an important part of the action's circumstances which contribute to the production of its consequences.
In applying UG, one must determine what the consequences would be of everyone's performing acts of the relevant sort; this in turn requires one to determine how many agents will have the opportunity to perform acts of this sort) But, as Sobel and Silverstein have pointed out, the question 'How many agents can perform acts of type A ?' can be interpreted in two different ways. 3 On the more natural distributive interpretation, the question means 'How many agents will have as individuals the opportunity to perform an act of type A ?' Since there are insurmountable difficulties with the theory on this interpretation, these investigators have urged a second, or collective, interpretation, according to which the question means 'How many agents can perform acts of type A collectively or together?' Clearly the answers to these two questions are different, for, when two people pass each other on the street without saying anything, both have the opportunity as individuals to be the only one to say 'hello ', but it is possible for only one of them to perform this act when they are viewed as a collectivity.
In a previous paper I argued that the distributive version of UG (on a Lyons-type interpretation) cannot be used by agents as an action-guide, since in many cases an agent who attempts to use it in deciding whether or not to perform an act of type A cannot determine how many agents will have the