This paper is concerned with the sources of variation in the earnings of American scientists over the decade 1960-70. It focuses on sources of variation over time as well as at a point in time. Earnings variation over time is decomposed into sources due to measurable variables representing an earnings function, a random effect individual variance component in the level of earnings, a random effect individual component in earnings growth, and a serially correlated transitory component. Maximum likelihood estimates of the implied parameters are presented and a method for obtaining GLS estimates of the earnings regression function, when the residual variance components are given, is explored.