States have used variants of these three approaches to structure the welfare to-work component of welfare reform for single mothers: (1) put education or training first; (2) place job searching first; and (3) create a flexible program that allows staff and participants choice in the initial and subsequent activities. Research on these program strategies 1985 to 1999 is unusually reliable because of the following: (1) it covers a variety of programs and almost 100,000 single parents (sufficient to reliably assess program effects); (2) it follows people for 5 years (long enough to determine whether an up-front investment in education pays off); (3) it is a real world measurement of what the three strategies produce; and (4) it uses random assignment (the most powerful research design). All three strategies increased single parents' work and reduced welfare receipt compared to what would have happened in the absence of the programs but did not increase people's income or have many or consistently positive or negative effects on children, except adolescents. Best results came from the flexible programs. Findings show there is no evidence to support a rigid education-or-training first policy; there is a clear role for skills-enhancing activities in welfare reform; the solution to low earnings is not in the training programs used in the past but should instead include training that fosters career advancement, integrates basic education and skills training, and engages local employers; support services should be provided; and, while well-designed welfare-to-work programs can increase earnings and reduce dependency, there are limits to the approach. (SLR)