An appealing, but invalid, derivation of the probability that at least one of n events occurs is justified, using a particular definition of subtraction of events.The probabilities that exactly m and at least m of the n events occur are derived similarly.