Although neural networks do offer a few advantages over some other nonlinear methods, in certain situations these advantages are difficult to utilize. However, many neural network applications in the social sciences are flawed in ways that obfuscate such effects. In this article, the neural network methodology is reviewed, some common flaws are pointed out, and a rather commonplace data set—dealing with school delinquency—is analyzed for illustrative purposes.