Corrective software maintenance, which consists of fixing defects that escape detection and manifest as field failures, is expensive, yet vital to ensuring customer satisfaction. To allocate and use maintenance resources effectively, it is necessary to understand the defect occurrence phenomenon in the field. A preliminary analysis of the defect occurrence data suggests that software defects vary in rate from corner cases, which may occur only once, to the pervasive, which occur many times. This suggests that the distribution of occurrence counts is heavy-tailed. Theoretical reasons and mounting evidence indicate that the distribution of defect occurrence rates is lognormal. We hypothesize that the distribution of occurrence rates is lognormal, and further hypothesize that the distribution of the number of occurrence counts follows the Discrete-Lognormal, also known as the Poisson-Lognormal. We confirm that hypothesis, using a variety of data from widely used networking software. We also discuss how the Discrete-Lognormal applies to subsets of defects, where subsets are formed according to year of occurrence, products, Orthogonal Defect Classification (ODC) Age, and severities. We use straightforward interpretations of the parameters of the lognormal to understand how values differ for different types of defects and for different software characteristics.