To determine whether leaving messages on answering machines would aid control recruitment via random-digit telephone dialing, a randomized trial was conducted during 1992-1994 involving 1,323 western Washington households with answering machines. For the experimental group, a message was left informing them about the study and promising a call-back; for the control group, no message was left. Leaving a message increased the response rate by about 20 percentage points (p = 0.002). More households were successfully screened for eligible controls, and individuals found eligible were more likely to participate. Leaving a message can help to improve response rates in telephone surveys.