PURPOSE: to study the sensitivity and specificity of automatic ABR in comparison to diagnostic ABR of newborns under risk in a neonatal hearing screening program. METHOD: one hundred and eighty six neonates were evaluated, 83 born at term and 103 were pre terms, 88 male and 98 female gender. Evaluation procedures consisted of transient otoacoustic emissions(TOAEs), A-ABR and diagnostic ABR carried out in the same week. RESULTS: from the 186 evaluated newborns, 156 (83.9%) showed normal hearing. Conductive hearing loss was verified in nine neonates (4.8%), seven bilateral and two unilateral. Cochlear hearing loss was observed in five neonates (2.7%), one unilateral and four bilateral. Central evidences were observed in eleven neonates (5.9%) and one neonate was diagnosed with auditory neuropathy spectrum disorder (0,5%). In four cases, we found a maturational delay in the first evaluation with normal result within one month. In the comparison of A-ABR with diagnostic ABR, high sensitivity rates were observed (99% and higher) in the identification of cochlear, conductive, central and auditory neuropathy. Specificity rate of A-ABR was 100% for auditory neuropathy, 75% for right ear and 60% for left ear cochlear hearing losses and 54.5% for right ear and 63.6% for the left ear central abnormalities. For conductive hearing losses, specificity was low (lower than 43%). CONCLUSION: A-ABR was efficient in the identification of auditory neuropathies with high sensitivity and specificity rates. However, false negative responses were observed in the group diagnosed with cochlear and conductive hearing losses, central abnormalities and maturational delay.