This study was performed to evaluate participation rates and characteristics of participants in a case-control study of infant leukaemia. Eligible cases were identified throughout the United States, and two controls were selected for each case using a structured protocol for telephone random digit dialling. The non-participation rate, once a household agreed to provide an address to receive a letter explaining the study, was 7.9% for cases and 19.8% for controls. When comparing demographic differences between participants and non-participants, parents of eligible cases that participated were more likely than non-participating case parents to be white (P = 0.01) and to be currently married (P = 0.03); parents of eligible controls who participated were more likely than non-participating parents of eligible controls to have a higher education level (P < 0.01), more children under the age of 10 years (P = 0.05) and to be white (P < 0.01). These data have application for future studies involving infants with a serious disease and the use of random digit dialling for the selection of infant controls.