The records of 94 consecutive developmentally retarded children with speech retardation and suspected hearing loss who underwent auditory assessment by both conventional behavioural observation audiometry (BOA) and brain stem evoked response audiometry (BERA) were analysed. In 54 children (57.4 per cent) there was good agreement between the results of both techniques leading to a clearcut diagnosis. In 22 children a diagnosis was possible only by the results of BERA as the results of BOA were inconclusive. Of the remaining 18 children, two groups could be identified whose results posed a dilemma. Group 1 (n = 7) consisted of children whose BOA test results differed considerably from their BERA results. Group 2 (n = 11) consisted of children in whom there was no discernible response by BERA while the response by BOA was either inconsistent (n = 5) or not elicitable (n = 6). The specific strategies to be adopted for hearing assessment in these situations are discussed.