Rain is not uniform in time and space in semiarid areas and its distribution is very important for the runoff process. Hydrological studies usually divide rainfall into events. However, defining rain events is complicated, and rain characteristics vary depending on how the events are delimited. Choosing a minimum inter-event time (MIT) is a commonly used criterion. Our hypothesis is that there will be an optimal MIT that explains the maximum part of the variance of the runoff, with time to runoff used as a surrogate. The objective is to establish a procedure in order to decide upon this optimal MIT. We developed regressions between time to runoff (T0) and three descriptive variables of rain. Our results show that the optimum MIT is 1 hour, which seems to be the minimum period of time required for water in larger macropores to drain and sufficiently modify the effect of antecedent soil moisture on the runoff generation process. Rain events are classified into three significantly different groups: (1) large and intense rains, (2) light rains on wet soil, and (3) light rains on dry soil. Intense rains produce most of the runoff, but there were significant differences between small events in the runoff generated. Of rain events, 63.75% are single-tip events, and many could be dew.