Part of the "conventional wisdom " on the subject of the size distribution of earnings is that earnings are approximately lognormal, but with an upper tail which is better described by the Pareto function. 1 The primary purpose of this paper is to re-examine these propositions, using British New Earnings Survey data disaggregated by occupational groups. We have several reasons for feeling that such a re-examination is timely. First, most previous studies have considered the distribution among all employees, while theoretical considerations suggest that disaggregated data might be more appropriate. We elaborate this point in Section 1 of the paper. Second, the method of estimating the parameters of the Pareto and lognormal functions has not always been given the attention it deserves. In Section 2 we address this issue, and also refer briefly to the related one of goodness of fit. The remainder of the paper presents parameter estimates for the Pareto and lognormal functions, using first aggregate data (Section 3), and then distributions within occupational groups (Section 4). Section 5 compares these two sets of results, and at the