A consensus has been forged in the last decade that recent periods of sustained growth in total factor productivity and reduced poverty are closely associated with improvements in a population's child nutrition, adult health, and schooling, particularly in low-income countries. Estimates of the productive returns from these three forms of human capital investment are nonetheless qualified by a number of limitations in our data and analytical methods. This paper reviews the problems that occupy researchers in this field and summarizes accumulating evidence of empirical regularities. Social experiments must be designed to assess how randomized policy interventions motivate families and individuals to invest in human capital, and then measure the changed wage opportunities of those who have been induced to make these investments. Statistical estimation of wage functions that seek to represent the relationship between wage rates and a variety of human capital stocks may yield biased estimates of private rates of return from these investments for a variety of reasons. The paper summarizes several of these problems and illustrates how data and statistical methods can be used to deal with some of them. The measures of labor productivity and the proxies specified for schooling and adult health are first discussed, and then the functional relationships between human capital and wages are described. Three types of estimation problem are discussed: (1) bias due to omitted variables, such as ability or frailty; (2) bias due to the measurement of an aggregation of multiple sources of human capital, e.g. genetic and socially reproducible variation, which may contribute to different gains in worker productivity; and (3) errors in measurement of the human capital stocks. Empirical examples and illustrative estimates are surveyed.