We consider the use of sigmoid functions for multistage detection in asynchronous code-division multiple-access (CDMA) systems. The sigmoid decision function for each stage of multistage detection is derived under the assumption that the residual noise which remains after the multiple-access interference (MAI) cancellation at a stage is Gaussian, It is suggested that the sigmoid function should be adjusted depending on the residual noise power at each stage. In addition, multilevel quantizers that best approximate the sigmoid function are designed. Computer simulation results demonstrate that multistage detectors employing these soft-decision functions perform considerably better than those with hard decision.