The origin of asteroid families in terms of collisional breakup is analyzed using the data by Williams (1979). The observed mass distributions of families and its implications are discussed, and the distributions of relative velocities are analyzed via a simplified one-dimensional approach, separating the entire sample of families into a few subclasses sharing some important 'physical' features. The possibility of reconstructing the three-dimensional ejection velocity pattern of the objects forming a family from the proper elements is discussed, and it is shown that this potentially fruitful method fails to provide relevant results because of the difficulties noted by Brouwer (1951).