To date, there has been a large bundle of literature concerning the impacts of Welfare Fmployment Programs on welfare recipients. This paper is to summarize their main findings and then review several important issues related to evaluation methodology. Based on my critical observations, I conclude that there is a clear indication of positive earnings and employment impacts across most of the WEP-based studies; however, the evidence is still diverse to some extent and is subject to several qualifications. In addition, the following conclusions are quite certain at this moment. An experimental design with randomization is the most favorable way to produce robust estimates for program effects; however, there are several limitations in performing randomization and the result is also generalizable only for the target population. on the other hand, the selectivity or selfselectivity bias may be produced in the absence of random assignments.