Morality Games 305 likely detected them and will likely levy sanctions. So the United States’s best response is to levy sanctions. Similarly, if the United States does not detect chemi- cal weapons, it expects France did not and will not levy sanctions, so the United States is better off not levying them. This result is useful for evaluating whether it is worthwhile to uphold a norm. The Obama administration was harshly criticized for threatening to go to war after the Assad regime used chemical weapons but not earlier, although the regime had already killed tens of thousands of civilians. The model clarifies that Obama’s posi- tion was not as inconsistent as his critics had charged: The norm against chemical weapons may be worth enforcing since it is sustainable, whereas norms against civilian casualties are harder to sustain and hence might not be worthwhile to enforce. Let’s return to some more of our motivating examples. Our model can explain why we define murder categorically: It is not possible to punish differently for dif- ferent amount of quality life years taken, but it is possible to punish differentially for a life taken. As with omission—commission, however, we do expect sadness or grief to depend greatly on life years lost, even if the punishment or moralistic outrage will be less sensitive. This is a prediction of the model that, as far as we know, has yet to be tested. Similarly, the “‘one-drop” rule is a categorical norm, so it can be socially enforced in an apartheid society. In contrast, consider a rule that advocates giving up one’s seat for someone with lighter skin. Since this is based on a threshold in a continuous variable, while it might be enforceable by a unilateral authority, it cannot be enforced by “mob rule.’ Other forms of discrimination, such as discriminating against the less attractive, or the less tall, or the elderly, all being continuous variables, cannot be socially enforced via coordinated punishment, and hence, we expect s