Several empirical regularities motivate most theories of the distribution of labor earnings. Earnings distributions tend to be skewed to the right and display long right tails. Mean earnings always exceed median earnings and the top percentiles of earners account for quite a disproportionate share of total earnings. Mean earnings also differ greatly across groups defined by occupation, education, experience, and other observed traits. With respect to the evolution of the distribution of earnings for a given cohort, initial earnings dispersion is smaller than the dispersion observed in prime working years.We explore several models that address these stylized facts. Stochastic theories examine links between assumptions about the distribution of endowments and implied features of earnings distributions given assumptions about the processes that translate endowments into earnings. Selection models describe how workers choose a career. Because workers select their best option from a menu of possible careers, their allocation decisions tend to generate skewed earnings distributions. Sorting models illustrate this process in an environment where workers learn about their endowments and therefore adjust their allocation decisions over time.Human capital theory demonstrates that earnings dispersion is a prerequisite for significant skill investments. Without earnings dispersion, workers would not willingly make the investments necessary for high-skill jobs. Human capital models illustrate how endowments of wealth and talent influence the investment decisions that generate observed distributions of earnings.Agency models illustrate how wage structures may determine rather than reflect worker productivity. Tournament theory addresses the long right tails of wage distributions within firms. Efficiency wage models address differences in wages across employments that involve different monitoring technologies.