It is shown how Bradford curves, i.e. cumulative rank-frequency functions, as used in informetrics, can describe the fragment size distribution of percolation models. This interesting fact is explained by arguing that some aspects of percolation can be interpreted as a model for the success-breeds-success or cumulative advantage phenomenon. We claim, moreover, that the percolation model can be used as a model to study (generalised) bibliographies. This article shows how ideas and techniques studied and developed in informetrics and scientometrics can successfully be applied in other fields of science, and vice versa.