The Auditory Response Cradle (ARC) is a fully automated microprocessor controlled machine that was designed for the hearing screening of full term neonates. In order to evaluate the ARC, 6000 babies were screened at a district maternity hospital over a period of three years. Every infant subsequently entered a three year follow up programme. One hundred and two babies (1.7%) failed the ARC screen (that is, they failed two ARC tests) and 20 of these were found to have some hearing impairment: in 10 it was severe (80-90 dBHL), in seven moderate (45-60 dBHL), and in three it was mild to moderate (less than 45 dBHL). In addition, of the 20 babies who failed a first test and were discharged before a second could be performed, two were confirmed to have a severe hearing loss; 79 infants failing the screen were cleared on further testing, giving the ARC a false positive rate of 1.3%. On following up all 6000 infants for three years, seven children who passed the neonatal screen were subsequently found to have a hearing loss. For two babies the aetiology was unknown but for five the hearing impairment was either due to a hereditary progressive loss or definite postnatal factors. Progressive and acquired hearing losses cannot be detected at a neonatal screen and this emphasises the need for follow up screens at other stages in the child's life. In this long term study the ARC has been found to have a high detection rate for severe hearing loss and confirms the practical possibility of using a behavioural technique for the universal screening of hearing in neonates.