Electoral registers are often used to randomly select persons for epidemiological surveys. The importance and the different types of inaccurate informations found in these lists were studied during a survey among an elderly population of 5,161 persons aged 75 and over in Tarn area. Information was found to be inaccurate in 13.7% of the cases (people who have died 3%, moved out of the area 10.5%, inaccurate date of birth 0.2%), this increased with age and with size of the parish. The overall participation was 65.3%. The corrected participation (with exclusion of inaccurate information) was 75.7% and was higher in rural parishes and where the help of local organizations of elderlies was available.